how did I beat Michael Gisler if his average power is more than mine and he weighs (much) less! Or Pinilla and Craig for that matter!!
I posted some thoughts on this a few months ago, which boiled down to the fact that you can produce more average watts but go slower as resistance is not linear, i.e. not directly proportional to speed (more like the square of the speed).
For example, if on the flat you do 200 watts for an hour you will average 200 watts but go faster than if you do 100 watts for half an hour and 300 watts for half an hour, still averaging 200 watts. This is because you lose more to resistance when doing 300 watts than you gain when doing 100 watts (assuming both on the flat).
The most efficient power to speed, even on a virtual trainer, should be to maintain exactly your average speed the entire race and never vary, the more you vary above (and below) your average speed the more you expend watts overcoming exponentially increased resistance.
Only in a vacuum on a frictionless surface would higher average power always mean higher average speed.
So one strategy is to push more on the slow sections and if need be recover on the fast sections, to reduce the variation from the average speed.
But you also need to avoid over extending yourself, so balancing a need to not exceed your sustainable effort level for long periods of time with the benefits of reducing the variation of speeds should give you the best overall outcome.
In the end it is not the average power you produce that counts, just the average speed!
Geoff