The energy wrt. vacuum of the occupied states is normally quite good for GGA, even though the band gap is off. As MGGA does not give usable total energies, it is not really an option in this case, as we need both VBM and total energy calculated with the same absolute reference.