Two pass variable bit rate will give a better result in theory, as you will get a higher bit rate in the scenes that really need it (pans etc.).
i.e. Say you encoded at a fixed bit rate of '7'. This would mean that static scenes would be encoded at 7 and heavily panned scenes would also be encoded at 7. Now 7 might not be enough for the pans and they end up pixellated and blocky.
With VBR a static scene might look identical at '3' and the '4' it saves on that scene could be added to the panning scene where it needs it, and the pan gets encoded at '11', and now looks sweeeet.
This is ultra simplified, but that is why VBR is usually better - The difficult scenes get a higher bitrate (less blockiness) than they could possibly get with CBR. If your software is crappy at analysis though, some static scenes may not get enough bitrare and look a little off.
Thats where the pro software is great, it allows you to set bitrates on a scene by scene basis if you are not happy with the automated process.