Comment on the inverse polarization effect in metal-stripe polarizers