The primary issue is that the coefficients are often only useful if you also have access to the basis functions, which is very difficult. Certain types of PDOS and population analysis are possible to do just having the coefficients and the overlap matrix, I agree, but a better way forward would probably be to specify exactly what you wish to compute, and we can find a way to do that (for us, it's quite easy), whereas exposing the coefficients etc PLUS explaining how to use them is - actually - a bigger job
