I tried using the do novo splice prediction to see if it could detect other splice variants in the androgen receptor. Though it correctly identified the exons, it couldn't identify any of the splice variants in the annotated file. So I was wondering if it is possible to input a custom annotation, or in my case, just add a few more junctions to cover the newly added AR variants from the literature?

Yes you can work with a combination of annotated and predicted gene models. The easiest way to do this is to use analyzeFeatures() for annotated gene models as explained in the vignette but with predict = TRUE.