I'm in a similar situation...I applied with a stated interest in particle phenomenology, but am interested in how particle physics, astrophysics, cosmology, etc. are all bleeding together as people try to concoct a fundamental theory. I agree with what dlenmn said about string theory, which is why I have tried to take an insane number of math courses so that I can feel like I really understand what I'm doing. However, I've tried to teach myself a bit of string theory on the side, and I've found that it gives you a really good intuition for how physics relates to geometry
. Things like the perturbation expansion in QFT (i.e. Feynman diagrams), the number of particle generations, even the Higgs mechanism can be represented in terms of geometry (the Euler characteristic of the string worldsheet, cohomology of various compact manifolds, and separation of several stacked D-branes, respectively), which I find absolutely fascinating.
Now, depending on your interests, you can go one of two ways. If you're more inclined toward the GR side of fundamental physics, you are used to thinking of physics (gravity) in terms of geometry, in which case studying string theory would probably teach you a lot. If, however, you are more inclined toward particle physics, you will probably be put off by the fact that the spectrum of real, physical particles seems to be the least important thing to most string theorists. In particular, I've found that trying to understand most of the particle content of string theory has exposed my lack of knowledge about gauge theories, supersymmetry, and spontaneous symmetry breaking. In that sense, it is much more important to study those things "concretely" in terms of particle phenomenology before attempting to study them "abstractly."
But in the end, string theory is just the quantum mechanics of relativistic 1-dimensional objects - all the higher math comes in when you try to apply it to the real world! Regarding your school choices, MIT has an amazing condensed matter theory group and people in the string community who are trying to apply their work to condensed-matter systems...see http://arxiv.org/abs/0807.1111
for example. They have relatively few people working on phenomenology, though, although they have Alan Guth, who is unbeatable if your interests lie more toward cosmology. Stanford also has a very strong string community, as well as particle phenomenologists like Dimopoulos. Caltech has many string theorists and cosmologists, but not so many particle phenomenologists.
Hope this helps...I think I'm leaning towards MIT myself, but maybe just because Stanford is 15 minutes from my house