Are you running in a replica set? If so then you can "smash the glass" :).
I had a similar situation and I ended up stopping the mongo process, made sure a new primary had been selected, and then deleted all of the data files. I then started mongo back up and it did a resync with the primary but into a nice clean DB structure (no waisted space).
This is obviously extreme, hence "smash the glass", and should only be done if nothing else works.
Cheers,
MJ