I've said it before and I'll probably keep saying it, I think trompe l'oreille (fools the ear) experiences from loudspeakers are mostly a white whale, something that is mostly impossible from the beginning.
The sound of an instrument in a space doesn't just come from its dynamic / harmonic content. It's also from its unique radiation pattern within the room. Loudspeaker radiation patterns are shaped differently than those of instruments. So even with infinitely perfect frequency and phase response, and whatever polar plot you may prefer, a loudspeaker generally won't sound quite like the instrument you are playing on it.
If you've ever taken a recording class, they show you where to put the mic(s) in order to capture a good sound from different instruments. Let me tell you that the radiation patterns of an acoustic guitar or cello are just not replicable with a speaker, unless you purpose-built it to have an F'd up radiation pattern like a guitar. At which point you have a guitar simulator in your room, not a normal stereo.
Like, sound comes out of the front in a couple places, plus the back and (to some extent) sides of a guitar. Capture all of those with several mics, fine. In order to (in theory reliably) make it sound really-real in a room, you need a speaker with transducers that fire more or less in the same directions as the guitar does. And you need to feed each transducer with the corresponding mic's signal.
I am not aware of this having been done EVER, even as an experiment, let alone via commercial stereo recording!
What we hear in recordings is what was picked up by microphones, which categorically sounds different than what you hear in a live space. If your speakers happen to make a mic sound like a live instrument, it's probably a synergistic coincidence, "broken clock is right twice a day, but in a good way" sort of thing.
The chances are maybe better if you record close, in a very dead space, mix it very dry, and play it back in a nice, live room. But this is generally considered to sound bad in terms of mixing. Not many recordings like this out there, guitar with no reverb at all.
When the natural radiation pattern is similar to that of a speaker, I think illusions are pretty achievable. So basically... close-mic'd voice and certain brass and woodwind instruments are plausible candidates. (if you play them only on one speaker at a time.) Electric guitar, totally doable since it comes through a speaker anyway. Piano, organ, orchestra, choir, full bands, drums in general... very iffy.
This is my ideological problem with trying to make recorded music sound "really real". It's a spurious illusion that probably involves some departure from true fidelity. It involves making some recordings better than they are, and by the same token, some worse. A truly accurate speaker will make recordings sound like themselves.
So I understand this is a compromise some people are willing to chase, but to me it feels like a gamble on a game I don't totally understand.