I disagree - I can't fully enjoy the performance of the star if it's a completely artificial construct, and a complete orchestra is reduced while the piano becomes a monster instrument spanning the entire stage.
I completely understand that your preferences are more of an ensemble recording approach than a hyper-realistic recording where the goal can be "bigger than life". I like both approaches depending on how good I find the particular recording to be, and as long as there are other aspects I find even more important like a convincing room sound to the recording, I'm usually fine either way.
While we listen to reproduced ordinary stereo recordings something really important is missing, and that is the visual clues. When we can't see the musician's performance in a live setting, the "cocktails effects" kick in and makes us able to focus on a particular musician and his/her instrument or voice. A substitute for that missing visual clues can sometimes be a more hyper-realistic recording approach, like simple stereo and panning effects that make things pop out more, even if that put the listeners in Keith Jarret's lap.
As I said, I also like the ensemble approach, but sometimes those recordings can sound a bit too distant. Microphones simply don't pick up the sounds as we hear them and without visual clues, the recordings can often sound somewhat bland. Stereo reproduction can do many different things, I think it would be boring if everything was recorded the same way.
And... I don't think it's common that Keith Jarret or others like him are telling the studio engineer how much space the piano should have on the recordings. It's more likely the person in the studio who recognize Jarret as the main musician in the band, and therefore makes that instrument the main attraction on the record.