It is absolutely clear to me from Winsplits statistics that long legs produce a considerably higher percentage of errors than shorter ones.
Yesterday at Drumlanrig http://www.elo.routegadget.co.uk/cgi-bin/reitti.cgi M21 had the rare treat of a real long route choice leg, which kept us talking and thinking long after the race was run. M35/40 had a similar, shorter variant.
Excellent though the leg was, and varied though the routes were, Winsplits doesn't show up any errors.