native datasets? And would a native dataset for refinement be compromised by friedel pair differences collected at an edge?
If you've got anomalous data available, you may get better results by using your friedel pair differences in refinement - especially if your diffraction stays in the ~ 4 Angstrom range.
This was originally implemented in refmac5D, and has been in the main refmac releases for a while (not sure about cns or phenix.refine - last I checked they didn't have a friedle pair refinement target).
Pete
