Hi Pavel,
Thanks for your reply. I figured out that my colleague didn't specify 1 or 0 and so the reflection file editor used 1 for test. I then used Aimless to transfer the FreeR set from his mtz and I think the new test flags are now 0, as I didn't specify it in phenix.refine and it automatically guessed the test flags are 0 (as shown in the log file).
As you said, if 1 and 0 are flipped, phenix will take 0 as test just going by the
fraction if the test flags are actually 0. But in this case, when I specify (or force) the test flags as 1, it seems that phenix.refine cannot flip it back to 0 automatically....or it will not be able to flip as I tell (or force) 1 are the test flags?
I guess what confused me before was that when I used Aimless to transfer the test flags, I didn't know whether Aimless knew that they were 1 or 0. If Aimless knew the test flags were 1, I was not sure whether Aimless would change them to 0 or the test flags would still be 1 (as I was going to use phenix.refine for refinement and I thought I had to specify it).
My understanding is that Aimless (which uses CAD to transfer the test flags) doesn't actually care 1 or 0 but only the column labels (which is FreeR_flag), but since Aimless is part of ccp4, the "new test flags" I have now becomes 0 and so phenix.refine recognizes the test flags as 0. Please correct me if I am wrong.