This means that you've changed the output type to amplitudes, but the labels still look like intensities (i.e. I, SIGI, etc.). If you are converting to merged amplitudes, then the output labels should be just F and SIGF (in that order). For what you're specifically trying to do, you will need to run it twice, once to generate F,SIGF (and FreeR_flags), and again to combine the output of the first run with the anomalous intensities.

> Also have hard time to convert them in command line, Is there any auto script that can work similar to scaelpack2mtz , which keeps all I+, I- and F , SigF, Rfree for refinement.

It's a little difficult to do this automatically right now. One reason is that Phenix doesn't deal very intuitively with MTZ columns (the units of data we work with - "Miller arrays" - are typically multiple columns grouped together). The other is that the reflection file editor is really meant to combine different data types, and doesn't do the sort of multi-step conversion that scalepack2mtz does. It would be trivial to add this feature as a simple black-box GUI widget (and command-line program) - I'll throw something together the next time I have a few hours free. The reflection file editor was really intended for operations like combining multiple datasets and manipulating R-free; merging and converting intensities was more of an afterthought.

There is one other key difference: scalepack2mtz also includes the French&Wilson data massaging in the program TRUNCATE, which guesses appropriate values for negative intensities. (I may be getting this slightly wrong; I never understood what this step does.) We don't have anything to do that in Phenix. Whether you actually need this is a matter of personal opinion, however.

-Nat