Hi Len, sorry for delayed reply - somehow I managed to overlook your email...
One of the groups here has run into a problem in the refinement of their structure or more specifically a ligand bound to the structure. Through a series of other experiments they know that a photo reaction occurs to the compound when exposed to high flux as found at your everyday synchrotron source. The exposure changes the bonding such that there are two conformations with a bit in between obviously. They collected 360 degrees worth of data and if the data is split you can model the two conformations depending which half of the data you use. They are trying to refine with all of the data and modeling of the ligand into the two conformations. In refmac they can model this with no problem but in Phenix the ligand adopts a chemically wrong conformation, sort of a halfway point between the two chemically relevant conformations. They have used Phenix to do similar refinements on other structures but it appears either the last release or the latest release is causing problems. They have started out with everything from 50% occupancy to shifting it towards on one conformation or the other and the ligand always ends up in the chemically wrong position. Data is in the 1.4-1.5 ang. range.
It's hard to tell what's causing this without looking at actual files. But I'm pretty sure it must be something easy to fix. If you send me an example: a fraction of PDB file containing atoms in question, and explain how you expect their occupancies to be refined, then I will send you back a working example and explain why it did not work before. I will look into it immediately once I have the files. Pavel P.S.: Please send the files to my email, not to the whole mailing list.