TLS Motion Determination http://skuld.bmsc.washington.edu/~tlsmd/
TLS Refinement http://www.ccp4.ac.uk/martyn/tls_research.html
Non-Crystallographic Symmetry exists when more than one identical object is present in the asymmetric unit. The operations relating the molecules do not have to be exact and non-crystallographic operations are not limited to crystallographic operations and their combinations (any combination of rotations and translations can relate two or more objects to each other).
Use of Non-crystallographic Symmetry in Protein Structure Refinement by Gerard J. Kleywegt
Notes from recent message on the ccp4bb
NCS usually good with 2.7A data;
TLS only to be used when model is complete;
You say Matthews suggest x molecules, but you only find x-y - doesnt this mean you have a very high solvent content?
Check the MR carefully
How are the x-y molecules related - are there dimer, trimers?
Submit coordinates to PISA, and to return the most compact packed structure
Does the self rotation agree with the above findings?
If so rebuild and refine a bit till R factors stop reducing,
then redo the MR with the larger unit
USE NCS at all times - certainly till the Rfactors have fallen below 30%
Once you think you have the complete and more or less corrected structure introduce TLS choosing the features to restrain together carefully. Maybe each complex belongs as a unit, or maybe a larger complex..