Hi Stephen.
I'm not sure this is really a defined use case for CMOD -- especially because it breaks loaded files into many little pieces. However, I have many customers that load 'BLOBs' (Binary Large OBjects) into CMOD.
There's no harm in trying this out and reporting back to us with your findings. Here's a few things I can think of off the top of my head:
I've loaded many extremely large files (32GB uncompressed) into CMOD, and not had problems -- that having been said, CMOD compressed the documents down to under 4GB. That 32GB file contained tens of thousands of documents, so each file was still handled in the way CMOD was designed to hold them: Split into documents, compressed, and packaged into 10MB pieces.
You'll want to increase the object size dramatically. I'm not sure what the upper limit for this value is (Default is 10,000K, or 10MB) but you should make this as large as CMOD will accept.
You'll want to pre-compress the file with another utility (bzip2 comes to mind because it's open source, multiplatform, handles large files, and compresses very well) and set CMOD's compression setting to 'Disable'.
All in all, it sounds like this would be better handled directly by TSM. TSM is built to manage objects like this (backups, archives, etc.), keep them in their original format, provide compression on the fly with the built-in hardware compression chips in modern tape drives, and migrate the data from one device to another.
Let us know what your research uncovers.