George is right about PWM being more efficient, but one should not underestimate the runtime gains from simply limiting the current with a resistor. I've gotten more than 24 hours of usable low level light from an ARC LSH-P with it's current limited to 30 ma by a resistor.
In designing new boards, changing the pulse width (PWM) is definitely the way to go. In retrofitting existing lights, a resistor may provide sufficient benefit without replacing the electronics of the light.
But I may be prejudiced in the matter; /ubbthreads/images/graemlins/smile.gif I'm still selling the hi/low spring assemblies for ARCs and similar lights.
Daniel