In this case I would simply read through the SET source code and compare the commandline arguments from SET, to the ones I would manually use in the Metasploit console to see if there are any differences, such as payload, settings, etc.
Play a bit with it, if you find a bug in SET, let the devs know
