The Vogl parameterization is rather rudimentary, it only contains sp3 orbitals, whereas Bassani (or Jancu et al., there are several authors on the paper) created a sp3d5s* model for Si and other semiconductors to make it more accurate. And it is - Vogl is well known to not describe the Delta valley of Si very well at all, and I think you will notice that esp. at high bias.
There have since been published improvements to the Jancu-Bassani models, in particular by Boykin, Klimeck et al., but they mainly adjust details.