1. The most important excitations are the waveport and the lumped port, and in my experience the second is even more important - but the solutions are/can be slightly less accurate. The waveport is actually a 2D eigenmode solver itself, so it finds (or tries to find) the correct excitation modes to apply. It can be used only on (perfect) conducting surfaces and bodies or on the outside of your model, where you apply radiation boundary or PerfE.
A good tutorial using the waveport is:
em: talk - HFSS Tutorial 1: Microstrip Patch Antenna
Normally, especially inside the model and for balanced excitation you use the lumped port. With that, you will have to define the port impedance yourself.
I also commonly use a coaxial stub (like a coaxial connector) which I build myself (I am sure you can figure that out yourself) using a waveport.
2. The most important boundaries are the radiation boundary (on the outside box of your model, if you model an antenna), the PerfE (perfect electric conductor) e.g. for simplified 2D structures etc, and the PerfH (perfect magnetic conductor), which can be used quite similar. Note that, if you don't apply any boundary on the outermost box or structure, HFSS applies PerfE automatically.
3. For the eigenmode solver you cant "decide" on a solution frequency, you can only choose the minimum mode frequency and the number of modes HFSS should find. For the driven mode, of course the solution frequency should be the one you want in the end, or, if you're looking for a frequency band, it sould be the uppermost frequency (fast sweep and discrete sweep) or the center frequency (interpolating sweep) - although it does not matter that much, normally (unless your doing UWB or something like that).
4. Note that on even on a perfect conductor, only the tangential electric field has to vanish, but not the normal components!