Some observations concerning the effect of impurities on the anatase-rutile transition