b-refinement

[Didier Nurizzo]

I have a dimer related by NCS. I use the NCS for the refinement in X-PLOR. 
So I have 15804 degrees of freedom and 37007 reflections (2.9 angstroms
resolution). In this case I suppose that I can refine b-factors (one for
the backbone and one for the side chain for each residue). But, the
b-refinement falls some b-factors at the minimum. And the rmsd of the
b-factors is very high. Did anyone know this problem and how can I
restraint bfactors to have good rmsd ? 

If anyone has suggestions or strategies for resolving the problem, I'd
be grateful.

Didier NURIZZO.