John,
I think the issue here is about real world, versus theoretical, performance.
For a well-designed motor, run within it's normal safe operating envelope, resistive losses should be very small, perhaps around 10 to 15%.
I have tended to simplify things by not emphasising the resistive loss element, as that isn't that important in terms of limiting practical working torque - maybe that was wrong.
As for the comment about those graphs showing the bottom end, in fact they show the whole safe working range of that motor. The stall current is many times the maximum current shown, but that's the case for any well-designed motor.
Any decent motor should be designed to have the very lowest DC resistance possible, for best efficiency. I've just looked at the specs for my 350 watt brushless model aircraft electric motor, it has a winding resistance of just 90 mOhms, for example, so even though it's maximum operating current is about 25 amps, it's stall current would be up around 150 amps at the normal controller operating voltage.
I don't have any figures for other bike motors, maybe they are much worse as you suggest, although the efficiency graphs available imply that they are similar. If they are as efficient as the motors I've played with, then there shouldn't be any really significant current reduction, and hence torque reduction, with a 33% supply voltage reduction, provided that the controller doesn't limit.
My real point was that a given current through a motor, at a given motor rpm, will give a particular torque, which is virtually independent of the supply voltage to the controller.
Jeremy