You were describing how a water brake engine dyno works. However, you have to realize that water brake dynomometer's are flawed more than inertial "spinning drum" dynos, here's the reason: Water brake dynos have water pumps, impellers, etc in the system made of rubber. To maintain the utmost of accuracy, these dynos have to be torn down and rebuilt anytime the dyno sets unused for longer than 2 weeks because the rubber develops a memory and flaws the pressure readings and thus the calculations determining the power at a given RPM, water brake dynos also need to be rebuilt periodically when they're used constantly due to normal wear that comes from rubbing against the pump housings. Failure to rebuild the pumps on a periodic basis in either condition causes a peculiar result: engines always show gains over previous tests, when in fact there is no true gain.
Inertial dynos can read power via the crank, or from the wheels. The advantage of an inertial dyno isn't in accuracy, it's advantage is that it can read and measure the power output of a setup during acceleration without pausing at various RPM intervals, which takes time and causes more wear on the engine than an inertial dyno. An inertial dyno's accuracy comes from knowing the diameter and specific weight of the roller drums, which [with proper periodic cleaning] doesn't change. Combined with magnetic pickups and a computer, an accurate pull is just a single run in second or third gear away. Now, for all intents and perposes, a chassis dyno measures how fast you can spin the rollers, that is, how much HP is making it to the wheels without being wasted in the tranny, turning the flywheel, or spinning the wheels.
Since we're discussing dynos, it should be mentioned that engine dynos will always show more HP than chassis dynos because drivetrain losses can not be accurately predicted, just as the "SAE corrected HP" for altitude differences are an educated guess at best on chassis dyno pulls.
Back to your equation, HP is torque applied over a time interval, and the way Hondas generate more HP is normally to apply that torque faster over that interval, which is acheived by reducing the inertia it is forced to accelerate if you can't make use of more fuel available in the cylinder to burn [stoichiometric ratio still needs to be observed]. So using your equation, are you still saying that if you reduce the inertia slightly and increase your acceleration more than necessary to maintain the same amount of torque you don't gain any power?
Of course, there is a limit which starts to yeild diminishing returns, but your comment of "but you don't have any more power. " isn't quite right since you're effectively decreasing inertia in order to accelerate the mass faster.
BTW, my statistics were done on a new inertial chassis dyno. We had to play
