A few observations:
Each object is about 3k (9.5GB/2.76M documents). It doesn't make sense to use large object support -- the individual objects are too small. Also, you said your compression is OD77, and the load message says it's using OD77Lite. You might be able to save some more space at the cost of a little more CPU (although I knows MIPS = $$$ on z/OS, so maybe that's a calculated move).
I'm not exactly sure why you're getting that message. It may be related to large object support (which stores a header in the object file describing the 'large object pieces').
It might be best to open a PMR on this and get an official answer, as understanding this at a deeper level would probably require access to or knowledge of the source code.