You don't mention some things like what Value the various civs have - differing values tends to lead to more wars. If you (or a single AI) are managing to claim most of them and only spreading 1 (or none), resulting in most (or all) of the AIs having the same Value this could reduce the war preparations. But this is probably not the problem.
I must say that I don't seem to have this problem as much as you do, although I think you are right that there are often fewer ships in the fleets of the AIs then there were in regular FF. Exactly why isn't clear.
Currently I'm thinking that a smaller ship count is a combination of 3 things in the way the game currently works plus maybe anther factor or two relevant to your options.
The first is something that you noticed - the composition of the barbarian fleets. They tend to get delta destroyers well before any player, and battleships at a time that is about the average for the players. This is because the instant the barbarian ship choices switch from the first era's list to the second era's list they get these ships. When does the list change? It changes when "enough" of the players have entered the second era - I think that is half of the players. The switch to the second era happens for each player well before the tech that enables the delta destroyer since it happens as soon as a player research any of: Shock and Awe (enables battleships), Streamlined Production, Vacuum Engineering, or Domestic Development (which is the path to the very valuable Empowered Citizenry tech - very valuable because of all the happiness stuff it gives since it not only enables the Sports Arena but the first player there gets to found the Religion value which has the special building that gives +1
).
Many AIs that have not already founded a value (and some that have) like to go for the Religion granting tech. (So do I, for that matter.) If they are actually in the running to get the tech first it put them into the second era via a path that doesn't add any new ship types other than the Delta PDS, which can make their systems harder to take but does nothing to help them attack anybody or fight off delta destroyers (or battleships). If they beeline for it they may not even have the Shield Generation tech which gives the ability to make Cruisers (which are generally the best ships to kill pirate delta destroyers since they can be built reasonably quickly and are slightly stronger than a delta destroyer) and which also leads to Shock and Awe. The net result is that many of the AIs are unprepared for the Pirate's ship list upgrade.
Another part of that same problem is that in the second era the number of Pirates also increases (it is, as I recall, proportionate with the amount of territory that is not inside anyone's cultural borders, and the percentage increases in the second era). So not only are the pirate ships suddenly stronger, there are also more of them.
This situation is likely to result in a lot of the AI's ships being destroyed, particularly since their use of squadrons is not as good as one might like (so they don't weaken the attacking pirates as effectively as they could).
A possible fix for some of this would be to delay the second era a bit. Perhaps moving the Streamlined Production, Vacuum Engineering, and Domestic Development techs from the 2nd era into the 1st.
The second issue that could be causing a smaller ship count is actually an improvement in the way the building production is done.
In regular FF when the AI checks to see if it can build a building it only checks the planet that is currently selected as its build planet (which is only ever changed in a rather random manner by some building choice AI code). This results in a lot of possible building choices being vetoed once a few buildings have been built on the current built planet - the building selection code in the DLL might want to increase production but if there is already a mining facility on the current planet then it won't be able to do so and will have to do something else (if it is the Python AI code that wants this, then it will try to pick some effectively random planet that doesn't have one as the current build planet - it checks a number of planets that are within the current influence range up to a count of 1 less than the current system population, checking them in the order they were added to the star system, and if it doesn't find one to switch to it bails out and lets the DLL try something).
In FF+, however, the code that checks to see if a building can be built will first check the current build planet but if that planet already has the building on it then it can check every planet in the system that is currently within the current influence range and switch to one that doesn't have it. This presumably results in a lot fewer rejection of building construction attempts.
Since FF+ is rejecting fewer attempts at construction buildings, it probably results in the AI building more buildings and fewer units than regular FF. This would result in the AI having a better economy and higher production, but a less effective fleet (until it starts running out of planets to build things on - at which point it should build more units).
Another related factor is they way I modified the AI to force it to build more squadrons (although it actually forces it to build more instead of buildings). They could be counted as military units which would result in a lower weight for the "build more ships" factor in the AIs decision making, resulting in a reduction in the production of other ships.
A fix for this could be to go through CIV4LeaderHeadInfos.xml and increase each leader's iBuildUnitProb value by a small amount, like 5 (or maybe 10). The current values are near the same range as the regular BtS values, from 15 for Lu Tianqu to 35 for Raul Colombo and Lukas von Reuther. BtS has Gandhi at 15 and Montezuma (and 4 others) at 35, but also has Mehemed, Napoleon, Ragnar, and Shaka at 40. As you can see, Lu Tianqu is the Gandhi of FF in terms of military - except that the value is possibly effectively lower due to the other issues.
The third issue, which is related to the 2nd, is that there are more things to build. More regular buildings, some national wonders, and some world wonders. This also gives it things to do instead of build units, and some of them take quite a while to complete, especially during the early game. It can take more than 20 turns to build a value shrine, or the Military Science Academy (usually the first non-shrine world wonder to become available). That's 20 (or more) turns where it could have build at least a less expensive building like a mining facility or mag-lev (or both) plus a unit. The fix for problem 2 is pretty much also the fix for this problem.
Your change to the star system generation may actually exacerbate the 2nd and 3rd problems by making it take longer for them to run out of buildings they can build. More planets = more copies of each building, which might equal fewer units being built.
The combination of these three things probably results in the AI tending to have a smaller military.
One of the "another factor or two" I mentioned is your choice of map size. The more room the AIs have to expand the longer they will tend to wait before getting into wars. The random distribution of the players would have an effect - if two start out close they will tend to fight earlier. If they are still focusing on expanding, it could at least aprtly explain the lack of a fleet.
Another part of it might be a consequence of playing above Noble difficulty in addition to the other factors. The number of pirates on Monarch is apparently something like 50% higher than on Noble (I never really play at anything but Noble - most of my games are test games, and getting wiped out, or even falling very far behind, is not good for testing things that happen in the mid to late game).
In my FF+ games, I generally see at least modestly sized fleets in pretty much every game, and often see some that are capable of conquering other AI's star systems (in most games I see AIs taking a few other star systems from other AIs, every once in a while I see them take a bunch) and I do occasionally loose star systems to the AI. I have even seen some of them try what is effectively a fairly early rush, starting perhaps 50 or 60 turns into the game at standard speed - AstroTech, the Forge, and the Brotherhood are the usual suspects for this, but they almost never succeed since they almost never bring enough invasion ships (the Forge tends to bring about one more than the others, presumably since its UU is a variant of the invasion ship, but that just means it often brings two instead of just one, or one instead of none - AstroTech seems to like to attack with a stack of 4-6 Missile Frigates, which is almost certain to fail).
The "not bringing enough invasion ships" thing is one of the factors that makes the AI less effective at invading than it could be (I'm thinking of trying out an adjustment to those units of giving them a small iAIweight value to force them to be built a little more often), the others are relatively ineffective use of squadrons and missiles and especially the fact that it never loads any of either of them onto ships. It would do a lot better on its attacks if it included a carrier full of fighters (or mix of fighters and bombers) in its invasion force, or even just put some missiles on its cruisers. Fixing either of those 2 unit carrying things would require some DLL changes to a couple of unit AIs, or at least some unpleasant (and probably not very effective) Python overrides to the unit AI.
As you can see, FF+ does still have a variety of problems - but it does usually play pretty well.
This week I may try a game or two with all of the leaders having +5 to their iBuildUnitProb values, giving the various invasion ship types a 3 for the iAIweight value, and moving the Streamlined Production, Vacuum Engineering, and Domestic Development techs from the 2nd era into the 1st.