Dear Michael,
Is there a particular reason that you are using SHELXL? CNS deals quite well with lower resolution data as it can perform torsional, rather than positional, refinement.
Remember with perfectly twinned data you are effectively halving the number of observations so your data will have the observation:parameter ratio of a lower resolution dataset. Is the 85% you quote within the 'allowed' or 'generously allowed' region of the plot?
When you say your chains are not identical, what do you mean? NCS is only appropriate where you expect the structure of the related regions to be identical.
Cheers,
James
