Changed the scoring so that is an error function: now it is the average sum square upper bounds deviation from the contact map. A perfect fit to the contact maps is err=0.
Cleaned up, refactored and documented.