Of course the refinement should proceed until some Free-R is achieved that
is commensurate with the R-merge of the original data.
We found once that the High-temperature model ended up mis-placed in the
unit cell. The cell had shrunk by about 2.5% in one direction on freezing.
The coordinates of the warm model were in Ang. units. When they were
transformed into fractional coords in the new cell, they were about an Ang.
away from where they should be. Ultimately the structure yielded R-free =
0.18 for data 10-1.6Ang.
