Question: how likely is it that they could reverse out a segment (using a known stolen file) that matches a section in the truecrypt image by brute force?
True Crypt also implements xor-encrypt-xor which means you don't only need to know where the data is but also to which blocks it was written to and what as int the adjacent blocks. Later versions of TrueCrypt implemented XTS which even made this more "complicated" to guess, especially when you tried intentionally to attack the block cipher by flooding it with data that you expect to be padded in a predictable manner.
I'm sorry but i for once would be really surprised if the US government or any other entity has actual cryptographic attacks against modern encryption standards that do not involve engineered back doors (And things like the ECC fiasco doesn't count since it's pretty unlikely that even with knowing the seed values the NSA could actually decrypt anything) into the cryptography software or the system in general.
They can and do of course attack the weak links and go after the users and their keys and password when they are vulnerable to be stolen.