Hi Christian,
if I understood the manual correctly I could refine the occupancy and set the B-Factor to a constant value, or vice versa.
you can do both, or either one, in any combination and applied to any selected set of atoms. So, yes, your understanding is correct.
Is there a possibility that phenix refine the B-Factor automatically to the value of the coordinating residues and adjust than the occupancy to a good value?
What you described should happen in practice, but since this is all done using restraints (and not constraints), the outcome can be approximate to what you expect (and cannot be promised for 100% sure). Indeed, the B-factors of coordinating residues (or any atoms close in space) are all restrained to be similar (using similarity restraints described here: http://www.phenix-online.org/newsletter/CCN_2010_07.pdf ; See article about ADP in phenix.refine.). The occupancies are refined too, if requested or if the initial values in input PDB file are between 0 and 1. So I guess all you need to to set up approximately good/correct initial values for B-factors and occupancies, refine them all and see what happens. Since refinement of occupancies and B-factors is correlated setting different initial values may result in slightly different refinement results. Let me know if you have any questions or need any help with this. Pavel.