[phenixbb] One adp per group gives very high variance between adjacent residues.

I playing around with ADP and TLS parameters for a 3.5 angstrom data set. When I use one ADP per group, the values can vary by as much as 100 between adjacent residues, which doesn't make physical sense to me. If I refine individual ADPs this doesn't occur, but R free is 4% higher. Is this a problem with my data, or something to do with the implementation of this feature? Or something else? Best, Kendall Nettles