Hi Dan,
Sorry I just saw your message but the protocol and options you described should be fine for detecting novel junctions with STAR.
Once you get to the MAJIQ builder stage there are a number of heuristics MAJIQ to determine if certain junctions are reliable and quantifiable (e.g. number of times the junction was seen, what fraction of samples in an experimental group had that junction, etc.) so it is worth checking out the help message and adjusting these parameters as you see fit.
-Matt