2) In simple cases, the UPS with subtracted secondary electrons directly depicts the DOS. In the supplement to this:
Article A TPD-based determination of the graphite interlayer cohesion energy
you find a direct comparison of UPS and a primitive theoretical DOS which was generated by putting Lorentzian functions over DFT orbital energies. As you can see, the agreement is already quite decent on this simple level. Deviations may be due to the absolute orbital energies, cross section effects and eventually the baseline subtraction also plays a role.
3) If you go towards materials with a higher crystallinity, a single UPS is no longer sufficient for that because angle resolution plays a massive role. In that case, one spectrum only depicts the dispersion relation over one point of the Brillouin zone, so you need a full ARUPS.