Matheus, Ian, Frank, Jonathan Thanks for all the comments. It is easy to calculate the amount of splitting of the spots once the cell dimensions of each component are known. For the P212121 case, with a-b twinning, splitting will be zero in the c* direction. The splitting will increase as h and k increase. The splitting would frequently occur in the phi direction and would not be as apparent if large phi rotation increments were used.
Ian's comment "In this situation the mosaicity becomes meaningless" presumably refers to the "mosaicity" produced by data processing programs. Yes, this will be rather compromised! However mosaicity will be a factor if one wants to measure the spots independently. The angular spread of the domains (if this is defined as mosaicity - terminology again!) will tend to merge split spots in to each other thereby providing a limitation if one wants to resolve them. This limitation is not absolute. In general the Bragg spacings of the spots will be different and they could be resolved as in powder diffraction. My not so hidden agenda for all this is to evaluate whether a higher resolution diffraction set up would allow one to measure the majority of the spots independently for pseudo merohedral crystals. Also whether this would be a useful thing to do (e.g. for twin fractions near 50% as in Ian's example). There are also rather more speculative possibilities for experimental detwinning for the 50% merohedral case. Finally, I think I have captured Yang's original thread. I hope you found the answer to your original question! Regards Colin