I just did a very simplistic calculation, and I get similar results to the PRL mentioned. Now, I didn't really take care about the alignment of the C60 relative to the crystal axes, so the symmetry point are clearly different (we see a direct band gap at our L, which would be their X).
As I said, this was quick-and-dirty; I used the Huckel model, non-selfconsistent. I tried both the Cerda graphite and the Hoffman basis sets, and apart from Cerda being slower, the results are qualitatively very similar.
Now, the really interesting part, at least in a historical perspective, is that the original calculation in the PRL took 6 hours on a pretty serious computer (and they had to stick to 2 k-points or else it would take forever). Mine took 15 seconds (Hoffman basis set) on a simple laptop, with 5x5x5 k-points!
If you want to be serious about it, one should do it self-consistently and take care to converge in k-points but also interaction range and mesh cut-off, since this is a rather strange "crystal".
