Strange, I've never come across the problem, but I can see from the patch that you're right. I'll fix this in a release tomorrow, and see if I can figure out why it hasn't affected me yet. By the way, I'm changing this method to derive the comp_root from the model class rather than the table, to keep closer to the paths() method in Maypole::View::Base. Should be a transparent change. Ideally I'd find a way of using the paths() method directly, but I think it spits out the comp roots in the wrong order for Mason.