phase issues, mic placement and mono compatibility

Adding two cents to this. Richard, indeed a real piano in a room has its sound reflected off the different surfaces. But whichever way you cut it, listeners have (a maximum of) two ears, to pick up all the complexity. That's the reason the ORTF mic practice Luc suggested works so well, it most closely imitates the way our ears work.

One thing to consider, that I expect might be related to the results both of you have been getting in your tests, is that the microphone placement modeling in PianoTeq seems to take place in an infinite/anechoic space. Meaning the microphones are only picking up the sound of the piano directly, unlike in an actual room where the sound also reflects off of all the surfaces, all those reflections eventually converging at the mics (or the listeners ears).

At least that's what the manual seems to suggest to me. Though it doesn't say so explicitly, it's implied when it recommends against turning off the reverb. My ears do sometimes feel like they're picking up room reflections even with the reverb turned off, but maybe the realism of the physical modeling is playing tricks on me. Perhaps Philippe can verify this? It would be interesting to know.

The alternative would be to have a physically modeled room with reflections for the microphones to pick up. Strictly speaking this would be a more realistic way to model a miked piano. My intuition is, however, that miking a piano in a completely anechoic space and applying reverb to the stereo recording can give just as satisfying results as modeling room reflections going into the mics.

Reverb does indeed become crucial though. Did you use reverb for your ORTF tests Luc? If not, that might explain your results not being satisfactory, a real life ORTF recording always has at least some room reflections in it. I suspect the type and amount of reverb would have to be dialed in ny ear to match the microphone placement. I'll try this out later and see what results I can get.

And as for the phase correlation issue, the reverb might influence that too, the time-frequency smearing could even out phase issues by adding complexity to the sound. That's also what happens when playing the sound through speakers into a room of course, the frequencies bouncing around the room and possibly arriving at your ears at different times and from various directions... this certainly explains this result you got Richard: "Perceptually, all of this is quite subtle."

I was initially surprised by that sentence, I would expect the opposite of "quite subtle" when reading nearly -1 on a correlation meter! But then I remembered what the meter is doing, and realized that it's measuring something that we never actually intend to hear. The measurement only represents reality directly if we sum the left and right channels into a single speaker. I expect if we actually did that we wouldn't describe the resulting differences in timbre as subtle.
Not saying this is not an issue at all, but I suspect that other factors play a bigger role in making electric pianos sound bad live. Not all sound systems are made equal... This is of course assuming the engineer is competent enough to realize that the piano channels needs to be hard-panned left-right. Without that there's sure to be problems.

Thanks for the food for thought!