> I don't care if
That's the thing: they aren't building it just for you. We can speculate about their internal requirements documents, but what's for sure is that they calculate a projected market adoption based on various factors like cost, market segmentation, power draw, heat transfer, noise etc. If it turns out that they project a bigger profit with a lower power limit than technically possible, that's what they'll do.
Personally, it would be nice of this were to apply to any CPU, GPU, SoC and even VRM and DC-DC regulators; let me worry about the dissipation, just pump out as many cycles as possible. But that's not really something that covers any significant market that Apple (or most multinationals) is targeting.
What I find much more surprising is that all of their M1/M2 deployments seem to be far fewer bin-limited; normally you'd get a crapton of SKUs for a SoC because they are so hard to manufacture, but Apple seems to get away with only 15 to 20. Perhaps this is also why there are much clearer power limits on most of their devices.