Seems like you're checking all of the obvious things, including the dreaded CR->CRLF conversion that the BLU manual warns about. As you suggest, I would try a different terminal emulator as my next step. (Also, is software flow control turned off? I don't expect it should matter, but it shouldn't be on...)
If that doesn't work, I might investigate the loaded data in Service Mode some more and see if anything looks suspicious. Maybe the first few hundred bytes are OK, but then the later ones become all zeros or something like that? The BLU manual says that the loader loads 53762 bytes, so addresses that do have loaded data should be $9C0 through $DBC2 I suppose.
If this investigation is inconclusive, you might try creating a binary file of 53762 bytes with different contents, perhaps an easily-recognisable pattern, and seeing if parts of that pattern are reproduced faithfully on inspection in Service Mode.
Note that the BLU manual's bootstrapping section also has instructions for selecting slower data rates by modifying the type-in Simple Serial Loader program---maybe one of these will result in a more reliable transfer?
Ah... fond memories of bootstapping BLU back in 2012, before I did the pad replacements for my Lisa keyboards. I wound up "typing" in the loader on the bare PCB with a "stylus"---that is, one of the few remaining good pads taped to the end of a pencil!
