I wouldn't recommend prioritizing followup by read-pair orientation. What I would recommend doing is prioritizing things by abundance of evidence. Having just one or two pairs with discordant orientations in a gene (or between genes) isn't as interesting as having 100 reads supporting a fusion/rearrangement (particularly if you have a couple reads indicating the break/fusion point). I would further prioritize candidates by biological relevance, if the fusion events accompany a phenotype of some sort (this may be a non-starter if you don't have much of a background in biology).