The point I think you're missing is that producing a constant 200 watts of power is a fantasy so any discussion premised on the notion of a constant power at the pedals is almost completely irrelevant to actual riding. As you said, people can't hold a constant power output to a good tolerance and one of the things that determines the power they can output is how hard they've been working thus far. So working 1-2 extra watts harder at the start of the ride could end up costing you 200-300 watts in your fatigued muscles at the end of the ride (and one should note that acceleration is always the first to take its toll and is thus more likely to have its effect multiplied than any other mechanical factor). After all, 75% of the power your body produces gets lost before it even reaches the pedals, so there's little point in considering the relatively insignificant factors of aerodynamics or rolling resistance until you've considered the body.

Your whole perspective is completely misguided because you imagine that somehow the power going into the pedals is independent of the mechanical losses. You're right to note that using HRM measurements is difficult and inconsistent, but there's really no other option and although using Kreuzotter analysis or speeds normalized by power meter measurements makes for simpler considerations, it does so at the expense of the most important real factors. The question is a hard one and although switching to an easier question makes things easier, it doesn't get you any closer to the answer.

Can't you see that a 5% loss calculated by Kreuzotter is meaningless in the face of the fact that there are 75% losses elsewhere (unaccounted for by Kreuzotter)? I agree that subjective reports are likely to be inaccurate and make for pretty slim evidence, but even something likely to be wrong or inaccurate is better than something we know to be wrong for sure (like Kreuzotter).