*** For details on how to be removed from this list visit the *** *** CCP4 home page http://www.ccp4.ac.uk ***
When you tried refmac, did you also try some different models for rigid groups using TLS ? I have seen several cases where there were reasonable displacements which could be modeled by TLS tensors, what led to a serious decrease on the R factors.
In my experience, even using just one TLS group per monomer has often led to a significant (>1%) decrease in Rfree, so it's not even always necessary to find creative ways to chop molecules into smaller groups.
If you haven't told REFMAC to generate all riding hydrogens, this could be one source of the problem - at 2.0A it will probably lower your Rfree by another 1% or more. CNS by default seems to ignore hydrogens, so I always see a couple percent improvement when switching from CNS to REFMAC (with hydrogens).
