Grinter,
I would be very careful when comparing atomic distances (or even considering them) using a 3.1 A data set. Take a look at what error estimates are for this resolution, given your R values ;).
Also, I second what has been said, that you should build a model that makes "physical" sense, and plus, you have a good idea of what your protein is like from your higher resolution holo-structure.
HTH
|