It may be that two different people have different results of being able to tell the difference due to physical differences, no matter the "effort" the put into listening, or training.
But most of the time that's intentional, as high frequencies you "can't hear" just waste bits that could be used on something more useful, or even cause noise from aliasing and harmonics, as no actual playback equipment is perfect. And extending the frequency range makes it much harder to design the circuitry. All for something you can't hear :P
It's another reason why high frequency (96khz) playback is kinda useless, or even make things worse, as the extra frequency range that gives cannot be heard by humans anyway, and just gives more opportunity for those higher frequency patterns to cause distortion. It may even be that people can "tell the difference" precisely because of those distortions. But that doesn't mean it's "better", indented, or even get the same result on different playback equipment.
The question is can you hear those differences, to which the answer is basically no, for a modern codec at high enough bitrate. "enough" is 128kbit or more in blind listening tests.
No one here is arguing that modern, high-bitrate codecs aren't much better at producing imperceptible artifacts. But 128kbps is absolutely not enough in blind listening tests, 128kpbs typically produces a ton of perceptible artifacts. You're just making that up.
To a “trained” ear that is listening out for the differences? Sure, those are perceptible. But those aren’t normal people.
AIUI there are some things that don't go away at any bit-rate, e.g. pre-echo.