Isn't this what we're all, those who listen to unamplified music at least, looking for? A greater sense of 'realness'. If it comes from a speaker that is less than flat, then maybe we should consider why that might be.
Nobody sits remotely close to where the microphones are placed in a concert hall. Are microphones picking up a different signal, with regards to frequency response/direct and indirect sound, than what we hear sitting in the audience? Likely they are (Linkwitz had something to say about this, too much power in the 2 to 4khz region when recorded, hopefully someone will come in with the link).
I understand the impulse behind getting a speaker that gives a flat response, but if recording techniques aren't a decent representation of what you hear in an actual live setting, then you may be moving further away from 'realness', rather than closer to it.
Accuracy to the recording is fine, but it is accuracy to something itself quite inaccurate. If we admit that the recording is a poor representation of the spatial qualities of a live event, then the recording itself doesn't become the yardstick by which things are measured.
The yardstick, for me, is whether something sounds more like a piano, a cello, a violin, an orchestra, and so on. For others, that never or rarely listen to unamplified music, I imagine their reference points are different.
It's an interesting point. However, how much comes from speakers, and how much from everything else? Think about Harry Pearson's old saw, the "absolute sound" being that you hear when sat listening from the middle rows of a medium sized concert hall. So, we get a concert pianist and a decent piano and listen from there. Now, go to the front seats and listen again. Now go to the back row or the circle, and listen again. Now go to a different, smaller hall and repeat. Now sit next to the pianist and listen again. Now visit the pianist in the room they teach in, probably with a different piano, and listen again. Now to to the recording studio, sit next to the recording engineer, and listen yet again. Now as a thought experiment, think about listening from where the microphones are....
You hear the pianist, playing the piano, every time. If the pianist has a distinctive tone, or interprets the piece in a particular way, you'll have heard that every time, won't you? But actually it gets very complicated and conflicted. We've heard it from different distances, with different pianos, in different acoustics. One thing I've picked up from talking to and listening to interviews with professional musicians is that they in turn react to the acoustic when performing. That's apart from the player maybe bringing a different mood or changing things to sound better as they play the piece each time. Yet, you will still know it's that player who is playing, even if you are listening to the recording later.
This is an interesting thing. My memory for the actual sound of a system playing back, to judge changes, works for maybe ten seconds and then fades. Yet I can spot a recording of Julian Bream playing guitar twenty four years away from having heard him play live, in a different environment (I'm referring to him because, well, he's pretty distinctive) from any of his recordings, when he is playing a different guitar, when the engineer has added reverb, through tape hiss. Now, let's listen to a decent recording through an abominable process, using the same speakers - say over DAB+ radio, using the dreaded 64kB AAC with spectral replication. Now I can still spot him from things like his wide use of different tones from changing right hand position, but to be honest most of the cues are gone. It's a muddy mess. Play a decent recording well over cheap bluetooth speakers, though, and he becomes clearly recognisable again. The speakers don't actually have to be brilliant, as long as they don't impose too badly on the playback. That's fortunate because it allows us to have working "entry level" speakers. It's why source first was important back in the days when the best we had were imperfect turntables, by the way.
I strongly suggest that for fidelity, it is more important to reproduce the musical details of the event than the spatial ones. Harry Pearson was wrong, and in a way so are you. Because you don't have what you are trying to reproduce. If you want to get closer to the real event as it was recorded, you need to actually reproduce, as best you can, what you do have. You need to reproduce as accurately as you can the information about what the player/s are doing and why, which means at least trying to preserve some of the information about the acoustic they are playing in, as well as the attack on the keys or strings, the control of the wind players, what instrument/s are being used, and so on. Attempts to make the recording into something else, whether by the mastering engineer over-imposing on it or by you building a "musical system" rather than an accurate one - or imposing overly strong early reflections in the listening room to try to create "greater realism" - work against the recording.
The yardstick can't be to produce what you heard from seat K25 and at least I think you've got that far. But (for me, anyway) neither is it "sounds more like a piano, a cello, a violin, an orchestra"... it should sound like the pianist, the violinist, the particular orchestra.