This would require the machine to CURRENTLY have the Truecrypt partition or file loaded. If the encryption key is already in memory, then the partition or file would already be accessible on the machine, wouldn't it?

The whole research paper makes no sense to me. Why bother freezing the ram and removing it to access the key to a file that is currently already loaded and unlocked?

Their "method" is useless when the machine is off, or if you haven't entered the password to it yet.