To improve your success with the practice exams, it depends first on what the problem is... that is, whether too much of the material is unfamiliar, or you have trouble solving the problems, or you can't do them fast enough etc...
If the material is unfamiliar, you'll of course need to review more. It's best to save the practice exams for after your review since the problems are much more instructive when you know the content and the challenge remains to solve the problem itself. I didn't even look at the practice exams before I was done a thorough review of my undergrad physics.
And yes, you'll need to learn on your own the basics of the subjects you haven't learned yet. My school doesn't teach stat mech and my quantum was taught incompetently. So I started from scratch with both of those and learned enough to understand nearly all of the related questions on the GREs. There's always a few abstact/formal quantum questions on the test, and a few relate to MB, BE and FD statistics and their applications so all that stuff is definitely worth learning, even if it takes a few weeks.
Also, note that you must thoroughly learn everything that's on the ETS website describing the physics GRE. Of course, it pretty much lists everything, but that's not what I mean... When they list small specific topics, such as compton scattering, x-ray diffraction etc... you must pay special attention to them.
If the issues are either speed or problem solving, you'll have to work on that in whatever way helps you, whether that's by plowing through Schaum's 3000 problems or sharing ideas with friends etc... Just remember that most of the problems do not take much computation, just a clever trick.
I've always wanted to work with fusion. I wouldn't mind doing other things instead. Ouch I know what you mean about liberal arts schools distracting you with nonsense courses. I'm at a large school that has an overwhelmingly humanities-biased core curriculum including required classes in "race studies" and art. Though I luckily had some decent profs in the humanities courses and they weren't nearly as painful as they could have been.
Indeed, the net physics learning available is nearly independent of the school, that is, an exact differential, except in extreme cases which we could call quasi-hectic, but the amount of nonsense endured depends on the particular program through which the student progressed.
I don't know much about condensed matter theory, but I'm pretty sure that is studied at nearly every major university in the country, so you should have lots of options.
Uh oh, I've completely forgot about Lagrange multipliers, haha.