This is exactly what I have read elsewhere, but I'm not sure it's correct. When a lower gear is used, the dyno will be spinning slower, so the power will be proportionally less, cancelling out that increase in torque.
This link is pretty good: http://www.superflow.com/support/cycledyn-theory.html
What I said in an earlier reply about the mass of the dyno being tuned to be about right for 4th gear (or a particular vehicle speed range) is definitely wrong - according to that link the mass of the dyno actually simulates a particular vehicle weight, so if the weight is a good compromise, meaningful runs in *any* gear should be possible, although the acceleration rates won't match on-road rates exactly, even if the dyno mass produced an exact match to the vehicle weight, due to traction and drag etc, I guess.
Anyway, next time I have a dyno done I'm going to ask them to do a quick run in 3rd gear. I bet the resulting power will be *less* than 4th gear, not more, because of the increased transmission losses in the lower gear. In 2nd gear, my AP22 accelerometer reports a maximum power at the wheels of 138kW, despite a very recent 4th gear dyno run reporting 168kW.
Greg.