So why do my ears tell me it sounds better when I use Dirac to correct my speakers up to 20 kHz?
Your ears could tell you anything. Your ears could tell you that $5000 interconnects transform the sound of your system to another plane. So what?
Why does almost all room correction software allow for correction above ~500 Hz if 'it definitely doesn't work'?
Marketing. The idea of it working well is compelling. Did you actually read with care the paper I linked in
my post that you are replying to? If you did, you would not be so confident that your above question is a reason that 'room correction of summed sound' is a good idea above 500 Hz.
The paper does not claim that summed-sound EQ above 500 Hz will
invariably make speakers sound worse. You would already know this if you had given it a good read, and wouldn't be misrepresenting it as you are. Rather, the paper demonstrates that such EQ turns the result into a lottery: the change could be for the better, for the worse, or null, but the whole approach simply doesn't generate the correct data to control which outcome occurs. So that makes it a really bad idea. And especially so when, in the case where the speakers themselves have outstanding direct-sound frequency response and excellent directivity control, then pretty much any EQ adjustment above 500 Hz is going to be detrimental. But if your speakers have pretty bad problems in their direct-sound frequency response, then the chances that the 'lottery EQ' will actually improve the direct sound improves...somewhat.
These are not subtle effects. The improvement in imaging and spatial clarity is unmistakeably real at mid to high frequencies.
Effects being 'not subtle' is not evidence that they cannot be the result of perceptual filters or bias effects, which can (and routinely do) create strongly 'not subtle' effects.
It may be that anechoic correction coupled with 'proper' (physical) room treatment is theoretically better at higher frequencies but the proof, as the saying goes, is in the pudding. Sometimes good enough is good enough ...
Sighted listening is not a pudding that proves anything about what is in the sound waves. There is no 'proof' in that 'pudding'.
PS - stop posting that people are telling you that you have cloth ears or are too dumb to tell good from bad sound. You seem to be revealing that you are not aware of psychoacoustic phenomena like the sighted listening effect, cognitive bias, placebo effects.... these are practical realities for all audiophiles to come to terms with. The whole idea that you, I, anyone here, is immune to these effects and can magically discern the sonic attributes without having to implement any experimental listening controls, is just not realistic. Yet such immunity is exactly what you are claiming to have when you assume that you are being insulted. Nobody is insulting you: they, we, are just mentioning the indisputable fact that we all mistakenly and naturally think we are much more immune to these effects than we really are.