Hi,

I'm working on a package that makes scripted calls to phenix.refine to generate a series of ccp4 formatted maps, and have run into an odd problem. Using the attached parameter file I get a series of perfectly normal-looking maps, except in the case of the filled 2mFo-DFc map (test-3_2mFo-DFc_fill.ccp4 - the second-last map generated). This comes out as a series of disconnected blobs (it looks rather like a mFo-DFc map, but it doesn't match the actual mFo-DFc map). When I take the resulting .mtz file and generate the filled map in Coot (with or without omission of Rfree flags), on the other hand, it seems to come out fine.  There are no error messages in the log, but I suspect the  default method to fill in the missing F_obs may be returning only the missing F_obs?

I've attached a small example image. Red wireframe is the 2mFo-DFc unfilled map at 1 sigma; green translucent is the 2mFo-DFc filled map at 1 sigma; yellow is the mFo-DFc map at 2 sigma.

The test structure I'm using is nothing special: 3WSY, picked more-or-less at random off the protein data bank. Phenix version is 1.9.1692, installed in Fedora 21.

Have I done something wrong that I'm just not seeing, or is there a bug here?

Thanks,

Tristan