automatically flip the unpinned faces to point down too, instead of trying to unwrap them in the other direction and giving bad results. If there's a mix of faces pinned up and down it will pick the direction with the biggest area.