Re: [ccp4bb] tls problem in C2221

[Pavel Afonine]

But you are correct that refmac should allow the residuals to go negative.  
  

This is what phenix.refine does when you do combined TLS + individual 
isotropic ADP refinement.

Pavel.