Possible loss of accuracy in curved (isoparametric) finite elements-comment on a paper by henshell, walters and warburton