It appears that P212121 twinned so that the apparent space group is P43212 or P41212 is more common than I had realized. This requires that two axes are approximately equal in length and that the reflections 00l with l=4n+1 and l=4n+3 are 'accidentally' missing. It is possible that in such cases translational NCS causes both the twinning and the apparent absences, and in addition it can disturb the statistical tests for twinning. These problems are likely to be more acute when the resolution of the data is relatively low.
George Prof. George M. Sheldrick FRS Dept. Structural Chemistry, University of Goettingen, Tammannstr. 4, D37077 Goettingen, Germany Tel. +49-551-39-3021 or -3068 Fax. +49-551-39-22582