Hi Enyuan,
I don't really know a rule of thumb for ionic bond v.s. covalent bond, but by default, the FEFF code will automatically choose reasonable muffin tins in most cases, and most of the time it overlaps them by ~15%, i.e. FOLP 1.15. In many cases, the overlap doesn't change the structure in the XANES a lot until you get far from FOLP 1.0. However, there are some special cases. For example, we have found that using FOLP 0.8 for hydrogen atoms seem's to work reasonably well in the XANES. This is because of the short bond lengths of H. In addition, there are cases where one atom is fairly isolated from the others, i.e., cage like structures, where FEFF can't overlap the potentials, and setting FOLP to something smaller than 1.0 is essential to make the code run. I would start by letting the code choose the muffin tins automatically, then change them and see how the results change. If the changes in the XANES are small, you probably don't have to worry about FOLP. Otherwise, this is probably a limitation of the spherical muffin tin potential. The best thing to do in that case, is to compare the DOS (LDOS) with a calculation from a full potential electronic structure code, i.e., WIEN2K, ABINIT, QUANTUM ESPRESSO, etc.
This can give you confidence in the FEFF calculation, or lead you to the best values for FOLP. You could also use the FOLP values to get the best comparison to experiment, but in some cases you may be fitting many-body effects, which can lead you to the wrong conclusions about your system.
Cheers,
Josh Kas