Hi Francis,
It seems that despite me specifying refinement.main.target=ml in my def file, phenix.refine seems to want to keep refining against HL coefficients.
If your input data file contains HL coefficients, phenix.refine automatically chooses to use them in refinement. To disable this, use "main.use_experimental_phases=false" parameter.
It seems overtly out of the way to remove the HL coefficients definition (both the label and file) from the def file as well. Am I doing something wrong then (i.e. does it keep going back to refining experimental phases for some valid reason?)
Run two refinements in parallel: with and w/o using HL, and see which one ends up with better statistics. Allow enough of macro-cycles for both refinements to assure full convergence (5-10). Pavel.