Dear modelling community,
thanks to Michael Buck for his reply, unfortunately the only one. I would assume that there are many modeling problems where it is inherently impossible to derive at a single "best" set of constraints and consequently a single model best satisfying these constraints. I can't believe that there is no general procedure to derive at some mesure of uncertainty associated with the coordinates of a given residue based on a) degree of conservation among template(s) and target, and, b) the local properties of the fold, and, c) variations in the restriants that follow from a) and b). Any helpful thoughts out there?
Michael: I'm not sure that I understand what a correlation between an uncertainty (statistical scatter) and error (systematic bias) tells me, or how the quanity you call "model error" is defined. In my case, the problem is a lot greater anyhow, as precise beginnings and endings of secondary structure motifs in target and template are highly likely to vary substantially (beta fold, low to moderate homology).
Cheers, margot