Dynos are "smart" and compensate for the gear change. You will lose a small amount of RWHP with numerically higher gears due to friction. The END.
So the dyno knows the amount of inertia in the vehicle? Possible, yes. As a coastdown test can derive that information. However, the Dynojet system does not use coastdown information to derive driveline friction and system inertia. If it did it would be possible to provide a realistic hp curve at the crankshaft.
What the Dynojet measures is the velocity of the drum (really the angular position versus time). Using the velocity it calculates the acceleration (or velocity change per unit of time as it does it in small time steps). It also records engine speed so it can translate the information into engine speed rather than vehicle speed.
Now what does measuring the drum speed mean? It means that anything that reduces the engines power translation into drum speed increase affects the measured power. The more inertia you have in the system (crankshaft, rods, flywheel, transmission, driveshaft, differential, halfshafts and tires/wheels) the less the power translation into the drum, and the more power stored in the rotation of those components. The inertia components can be divided into those that are driven via road speed (tires/wheels and half shafts) and those driven at crankshaft speed (crank, rods, trans, flywheel and driveshaft). The reflected inertial components (upstream of the diff) will store energy at a faster rate de to the engine revving quicker. The difference reduces the hp reading accordingly. This is why a lighter flywheel will provide a power increase, as measured on an inertial dyno, and not on a steady state dyno. The driven inertia will be at a lower overall speed, but will be accelerated at a higher rate. Their impact on the measurement is pretty much negated as a result (that due to a different gear ratio). But, just like a lighter flywheel increases the measured power, so will lighter wheels.
What happens with a reduces gear ratio is the drum is accelerated faster, but at slower overall speeds. While doing so the reflected inertia components are storing energy at a higher rate and reduce the measured power.
Note the differential will also have higher friction with increasing ratio. that is because the reaction forces go up, and thus the bearing loads go up. My expectation is that this is a fairly small percentage though.