I tried to reproduce your results obtained in sorghum by using the full-length RoIs from control and PEG treated that you uploaded (https://zenodo.org/record/49944#.WOSWKdLyjIU). So I parsed those reads with an identifiable 5', 3' and poly-A tail from both files and concatenated them. Then I ran TAPIS with the default parameters. However, when I compare these results with the transcripts inside the sorghum_isoseq.gtf file that you uploaded (http://combi.cs.colostate.edu/sorghum/isoseq/), my output is much bigger. Even though I'm obtaining almost the same number genes that you did (I found 14,531 genes), my output had a much higher number of different transcripts: 48,261 versus those 23,667 transcripts that you found.
So my question is if you did any further filtering after the final output that returns run_tapis.py . Should I modify any parameter to get better results?