See... you're simply reading the dynos wrong.
Dynos interpolate engine torque by measuring the force at the wheels, then back-calculating engine torque by comparing wheel speed to engine speed.
Those readouts
are not what the dyno is actually seeing, but the dyno's interpretation of the results, translated for the operator.
The raw numbers do not look
anything like that. Raw wheel torque or force numbers plotted against road speed look like this:
Or translated into acceleration potential, thus:
So, yes... third gear should pull harder than fifth at any speed where the two gears intersect (unless you have a really weird powerband). If gearing had no multiplication effect, you would NOT be able to do a wheelie in first gear, no matter how hard you tried.
-----
Furthermore, I think you're confusing terminology in your example. What you're describing being built up as the bike goes down the road is momentum, which is
force(mass) x velocity
or
force(mass) x (distance / time)
Yet you're labeling it "horsepower per second", which is
(force(mass) x velocity) / time
or
force(mass) x distance / time2
Which isn't momentum.
It's best if you simply drop engine horsepower from the equation completely, and focus on wheel torque, which you can extrapolate from a dyno plot and gearing data. Seen this way, it's pretty obvious that it's not built up momentum that is slowing down acceleration, but decreasing wheel torque due to lower and lower gear multiplication at high speeds.
-----
If you use a rocket instead of a drive wheel and drive your bike in space, you will be able to maintain perfect acceleration no matter how much momentum you put into the bike.
Well, at least until you experience speed-of-light related relativistic effects... but basically, a build-up of momentum is NOT what is slowing down your acceleration here on Earth.
