Ratio is not important, as it doesn't reflect what you think it does. Power is comprised of soldier count, certain techs known, and even particular buildings that have been put in place. The most common distortion of an AI's power is their navy, as they LOVE LOVE LOVE to spam ships on any map where they have coastal cities, and ship units can boost it quite a lot (such as when they get Combustion and start stacking up steel-hulls). An AI also never brings all it has to bear on you, as they always leave some count of defenders in all their cities (which can be up to 4 units or more), another artificial inflation of their capacity to make war using a simple single number.
The point is, a much more useful metric is the general military tech standing of your target (if they have longbows and you have Cuirs, well, bad for them) and direct comparisons of unit numbers in the immediate area -- even if an AI has 20 cities and some 80 longbows across their empire, they won't have all 80 of those in the one spot you choose to attack. Especially if you attack swiftly. The human player is MUCH better at mobilizing and maneuvering than the AI, who plays very reactionary.
A better evaluation of when you are ready to attack is whether you think you have enough units in your attack to make the drive, or to kill the target's assembled stack -- both of which you need to directly observe the AI for. Don't rely on the power ratio, use your eyes.
Do many of you wage limited war outside of these circumstances very often, or is aiming for total conquest most often the best way to go?
Total conquest is often a pretty lofty target. Typically in my games the goal of a first attack is to gain land, so capping the first target as long as I get several nice cites out of it is an acceptable outcome. Even better if you can cap them with a border intact between the other AIs to serve as a culture buffer to incoming enemy units in the future. If I am powerful enough to outright wipe them in the first attack (often the case, if you choose a proper target), it can just be better to take all of them to gain as much as possible and get ready for the future.
Once you have "enough" land and are ready to roll on to the next target, it's most often just cap ASAP for me if the idea is simply to win the game as fast/easy as possible.
In those situations, do you typically raze those cities or occupy them? I guess if you remove them from the map, you're foregoing at least small sources of yield, but they also seem like they could be liabilities to defend if you're just trying to force capitulation and won't end up being able to work their whole BFC anyway.
If you are simply trying to hamper an AI in late game (this quote is in respone to taking down runaways in Modern era), there isn't much incentive to keep anything, and razing is a very good way to actually hurt them...they can't take it back, it only takes one unit capturing to raze it and the rest of the army can stay together for survival/damage potential and move on, and they'll never be able to account for the loss of the fully mature city, even if they immediately resettle the location, in a short timeframe. Which by that time you could chip away at them again.
There are also more extreme cases like stopping a Cultural or Space victory, where you just YES raze the capitol or approaching legendary cities. Such a set back is devastating to the AI.