I assume you are using standard roller chain and different size sprockets to achieve a change in speed?
I usually use 60% efficiency as a 'rule of thumb' to get back to the required motor power. I find the tables very wide ranging so you don't have to be fanatically accurate with this sort of thing.
I think 60% for a roller chain drive is way too conservative. I think the riders in the Tour de France would be rather upset to learn that 40% of their efforts were being wasted in heating up their chain wheels! Freshly lubricated roller chain is typically around 97 % efficient at least, but it drops off after the initial lubrication. I'm sure someone has studied this from first principles, but it wouldn't be too difficult to do an analysis if it can't be located. I would guess that it would be highly dependant on the coefficient of friction used (ie lubrication), and the ratio between pin size and wheel diameter. Silent chain has more of a rolling action, and should be even more efficient. Given the coefficient of friction, I might be able to come up with an approximate formula if you are still interested and no one else responds.
