This is a bit puzzling.
Sticking to point groups:
P3 &  P6 are sub groups of P6/mmm so data which merges in P6/mmm will always satisfy P3 and P6.
And twinning in P3 or P6 will make the data seem to have higher symmetry.
Four way twinning is unusual, but possible of course.

But if you really have twinning, you should see it indicated in the pointless/aimless plots, or via Xtriage.. If those tests do not show it and you dont have a non-crystallographic translation twinning is unlikely.

Your choice of SGs seem puzzling too.
 If the data is P31 (or P32) then you should see intensities for l=3n, and absences for l=3n+-1
If the data is P61 (or P65) then you should only see intensities for l=6n.
If the data is P6322 you should only see intensities for l=2n.
What are the absences along 00l?

Eleanor






On 29 July 2013 18:37, Jeffrey D Brodin <[log in to unmask]> wrote:
Hi everyone,

I have a dataset that's been giving me some trouble and wanted to get your ideas on the best way to proceed. The data extend to ~2.6 Å and scale integrate/scale well in P622. According to Pointless, the symmetry is either P622 or P6322, however, neither Molrep norm Phaser finds molecular replacement solutions in these space groups. If I run Phaser and let it check all other space groups it finds a solution in P6122, but I can only refine it to an Rfree value of ~35%. The data also scales well in P3 and P6 and I get molecular replacement solutions in space groups of either P31 or P61. However, the Refinement again stalls at an Rfree value of ~35%. I tried adding twin refinement in Refmac and it seems to be helping; The R/Rfree values dropped to 20.9/27.3 and the map appears much better compared to the previous refinements. The refined twin fractions are 0.26, 0.24, 0.24 and 0.25. Am I doing something wrong with the refinement or is this a legitimate way to treat the data. Thank you in advance for your help.