Friction IS accounted for. Every piece of machinery has a "power cost" associated with it.
I personally think that the fault lies in the pump ITSELF. It's power-requirements are far too low given the amount of water it can move. A pump's flow-rate is limited ONLY by how quickly its intake is replenished. For 10 power, that's a bargain! Gear mechanisms use 5, for comparison.
The problem is that falling water produces more energy than it takes to raise it to that height. If falling water produces 90 power (discounting all friction and what nots) then raising that same amount of water to the top of the wheel again should always cost more than 90 power (unless you raise it infinitely slowly, which means it will never get there). I think the loss is in enthropy. At least, that's generally what energy loss through reactions is attributed to in chemistry.
Technically, enthropy is just the disbursment/degrading of energy and matter forms to unusable levels/states. As for the falling energy not equaling the raising requirement, is because, for one, not all the energy of the water falling is transfered into the waterwheel. If that was the case, it would stop falling once it left the water wheel. And you would also have energy loss to friction and sound as well, especially sound. Waterfalls are noisy.