I tend to cottage my capital and farm everywhere else. The capital generates a lot of beakers as it can easily build libraries/universities/Oxford and the other cities I use to grow and whip out units to fight.
This is my most common approach too. It's much more important to run the slider as high as you can than it is to have a lot of commerce. Cottage cities take forever to mature and until they do they are essentially dead weight that barely pay for themselves and produce practically nothing for your empire.
At other times I take an opposite approach and build no cottages in my capital and then probably have a cottage city somewhere else solely to cover my expenses. Occasionally I'll cottage spam extensively, but I have only done that a couple of times.
I will generally keep captured towns, however. After the AI has invested all the time and cost of seeing them to maturity towns are well worth having, and by then I will already have a very strong core production based empire.
You could also say that by favoring building farms over cottages, you have put yourself into a situation where you're more likely to rely on the draft. A cottage-based economy would have a better tech rate and might not need to draft so heavily. In other words, you're in a self-reinforcing situation here.
This is something of a fallacy. Cottages are only stronger later in the game, by which stage you might already have won if you had chosen a different approach that gives faster results.
In any case, I would say that a commerce rich empire is usually a production poor one, in which case the draft is more necessary. A farm, mine and workshop empire has no difficulty building enough units. Universal Suffrage does change the complexion of this, of course, allowing cottage empires to build things pretty decently.
regardless of what the improvement on the tile is. If you need more food, put a farm down; if you need more 
