Looks like I've underestimated some games I've played to Astro and said "Can stop playing here, I'll win anyway". In other words my position in the Charly game is probably not quite as good as I originally thought it was (winnable but not anywhere close to easy or "won") I've practiced some more and when I don't have a super-weak target to attack (like Joao in the Boudica game or Willem in the Toku-skywalker one), I found it very hard to build a decent army before DoW. So I gave it a bit of thinking.
What do we usually do? Reach early Optics with 4 crappy ungrown cities that have no infra. Then settle the rest of the island after Optics. I'm not sure what happens in your games, but in my games the newly settled cities almost never contribute by Steel time, so I have to wait quite a bit after reaching Steel before I can launch an invasion (in most cases). This does often lead to overwhipping and massive unhappiness... bad. It's A LOT better to attack @1100AD after reaching 1050AD Steel than to attack @1100AD after having reached 900AD Steel (also because it's much more common to trade for Rep parts @1050AD than @900AD)
The answer to this problem? Assuming you're not attacking a total pushover, you want to start the build-up ~20 turns before you reach your main military tech (unless you have very high food and a great happy cap). This means you will often need every city to have forge and barracks by Astro time or only a few turns after. And this means that your last cities should be settled at least 10 turns before you reach Optics (unless you can develop them extremely quickly ofc). An ocean-fish city settled when reaching Astro is completely useless..
Of course, this all depends on your commerce/production situation, but I'm now 100% sure that we're too focused on both getting early Optics and on reaching very high bpt ASAP. Isolation is not about settling 4 cities, reaching 1AD Optics and calling the game "won".. there's so much more to it...
Another big advantage of spreading the buildup over time is that you can later take full advantage of drafting: you will already have a solid unit base & low whip unhappiness, which will allow you to keep teching while drafting cheap Rifles.
Yesterday I felt ready to start that Kublai game, but now it looks like I'll need some practice first, otherwise my winning streak will most likely go "kapoof"
Edit - In a "normal game" we tech to a huge military advantage (Cuirs vs longbows or similar), whip everything to the ground and snowball -- so the research drop doesn't matter. In isolation the 2 differences are that we don't have a big military advantage (if at all) and that we can't have a big production burst -- no snowball. In a "normal game" it's common to use a lot of bulbs, in iso we barely use any, and no one questions that, aren't we doing something wrong? Considering that Chemistry and GP can be bulbed, isn't it better to simply stay isolated for much longer, not trade for anything and attack only a few turns after first contact?