Despite a good fit : model can still have problems : wrong chemistry was defined and structure not physically reasonable.
In this case, GSAS refined model was used to try and determine twin laws on the single crystal data using “Jana” for refinement and Simon Parsons and Bob Gould’s “Rotax” (Fo/Fc) twinning software; and pre-defined twinning models.
Refinement in R3 then C1 with and without various twin models gave a final R factor of 4.6%, untwinned in R3.
Using the powder refined model as a starting point, it would seem that “Jana” was able to minimize on the single crystal data in R3 (and C1) where other programs possibly failed.