In simple terms power = torque x rotational speed and this does not necessarily follow a linear relationship i.e. if the speed doubles but the developed torque say halves then the power produced is the same as before. I wonder if your power testing managed to develop a set of Power / Torque versus speed curves for the mill. This might have helped established the optimum speed at which the mill should run when in production.
6HP is not so different from that provided to a small portable band saw and so this sawmill should be able to do some useful work. What happened when the saw got to the end of its travel and hence had finished absorbing the power produced by the water wheel. Was there a general speed up in the reciprocation of the saw or did you have a flywheel momentum storage device to help smooth out power demand fluctuations ?