Fixed bug: now PhiPsiAverager behaves like GraphAverager:  consensus is on templates agreeing out of potential templates (non-gaps) instead of out of total number of templates.