On the choice of gauge for approximate quantum mechanical wavefunctions