Yeah, it's quite possible. I have two USB-MIDI cables, MIDITECH MIDILINK-MINI and one unbranded. The latter one is about unbuffered.
MIDITECH's one has buffer of ~128 bytes. So, basically, both are unusable under Windows with the default class-compliant driver. Unbranded cable is unable to send sysexes at all as the third byte and further are stripped from the USB packet by the driver. Surprisingly, but MIDITECH's cable also screws sysexes of length > 128 bytes while sending. But it works perfectly on _receiving_ any sysexes. The unbranded cable
completely sucks at receiving sysexes as it is unbuffered.
But under Ubuntu the unbranded cable works just fine (I mean sending sysexes of course, heh). So, I write a quick driver for Windows using libusb to emulate such behaviour. Now, the unbranded cable sends sysexes just fine but _cannot_ receive sysexes at all. Your case, Mau1wurf1977, could be very similar.
Well, to close this cheap&screwed-USB-MIDI-cable-related-issue. I also wrote sysex splitter (for Windows) which receives MIDI events and re-sends them to the destination MIDI port splitting large sysexes into several small ones (of course, in MT-32 sysex format only). So now, both my cables works more or less.
If someone wants to look at the stuff mentioned:
https://github.com/sergm/munt_devel/tree/master/MIDI_stuff