Okay looking more carefully at how much signal range you can get onto a recording. Let us say you'll record 120 dbSPL. Okay one of my quieter microphones has self noise of 4 db. So you have 116 db left right there. One of my better microphone preamps would add 2.5 db to that in its operations. So we are down to 114.5 db overall range of the signal. I'm leaving out the nasty details, but could supply them if needed. Now to get the levels to add up best possible I'll actually need a bit of gain to get that 120 db SPL peak to be at though not over 0 dbFS. I'll end up with about 109 db overall once that happens.
All of that is something that can be captured by well dithered 16 bit if one is careful and does no superfluous processing. Recording in 24 bit gives a little headroom and makes it possible to have the final result stuffed into dithered 16 bit. Now one might thru close miking and making stars line up perfectly do a little better than that. This microphone in the example is quieter than most in self noise, but it also can record sounds louder than 120 db SPL at low distortion. The ultimate limit is going to be the microphone pre at 128.5 db range if one were to somehow manage it, and in this case the other factors will still limit that to 114 db.
Now this ignores ambient noise at the recording location. It ignores the ambient noise at playback. Chances are you loose at least 20 dbSPL there, and even if sticking to frequency bands where our hearing is most sensitive you'll still have 10 db SPL ambient noise. So the potential to exceed the dithered dynamic range of CD is right on the edge. Maybe so, maybe not, and in most cases simply not possible. Few, few, few recordings will manage it.
So dithered 16 bit a limit on possible recording quality due to reduced dynamic range? Theoretically barely possible by a few db at the most optimistic. In the world of available recordings it probably is not the case anywhere close to 1 recording out of 1000.