cuffdiff: high-confidence transcripts?

[mrfox]

Hi all, i have a question about the "high-confidence" transcripts which are use as the input GTF file to cuffdiff. My understanding is, we should apply cuffcompare on the transctript.gtf files and select only the transcripts from stdout.combined.gtf with "j" or "=", which are either known or novel ones. The DE and splicing analysis will be focused only on these selected transcripts.

Am I right? Thank you.