Just a quick thought that might be a workaround solution within the PDB format restrictions: In NMR structures there are often multiple conformations stored in different MODEL sections - this should in principle be possible for x-ray structures as well, shouldn't it?
Of course this is a waste of space, as you have to duplicate all atoms and not only the ones that are included in multiple conformations, and not all features might be handled by validation tools, but with the limitations of the current PDB format there are not many ways to encode this anyway. Best regards, Thomas