Sending the template structure will give you more information about how are the model actualy because you can take also into account the crystal symmetry that most of the cases is different between the template and the real structure.
Note #1: This server was designed for the evaluation of models built by homology, hence we expect that the degree of similarity between the "model" and the "real" backbone will be high, and we have a cutoff for residue equivalence of 7 Ångstroms. If you have used a program that does not conserve the "real backbone" or you have inserted a new loop, and you have a difference of more than 7 Ångstrom between two equivalent residues, we will not take those in to consideration.
Note #2: Our superposition algorithm always tries to do its best, but if you have a model of only one domain of a multidomain protein, please, chop the "real" PDB file, and submit ONLY the domain that you are trying to compare. This is specially critical for proteins that have two or more similar domains, e.g. multiple IG like domains, and you have the model of only one of them. You can perfectly end up getting the wrong superposition.
We are using some special modules written in WHAT_IF that will automatically superimpose the equivalent structures, and then calculate a number of parameters like the RMS and Linear deviations as a function of atom class (alpha carbon, sidechain, etc), secondary structure elements and surface accessibility among others, followed by a summary on each of the secondary structure elements and then it will point out the 10 worst modelled residues.
If you have detected any error, or have any question or suggestion, please send an Email to Gert Vriend.