In my understanding of the subject this is not entirely true - though the source of original vibration may be relatively small, it is rarely a point source. Natural sound sources also have secondary sources of sound radiation. E.g. on a piano the bat hitting a string will be the main source (still not a point though since the whole string vibrates in a complex way) but the body of the instrument will also resonate and radiate sound. Here's an example of some investigation into this that I was able to find quickly.
Real instruments / sound sources typically also have complex sound radiation patterns that are probably futile to try and mimic with loudspeakers - I'm not even sure that would be a valid design goal - given that the full 3D soundfield of the instrument was never captured in the first place when recording it.
That is not meant to say that coaxial loudspeakers are not good or better - just that I don't believe we can directly extrapolate from natural sound sources when speaking of speakers (pun intended ).
how about we go one step further up the chain? it's very unusual to record an instrument with more than one mic, so to reproduce a single microphone recording with more than one point source / diaphragms is not really as exact is it?