I think this is a subjective decision. Although people call certain transcripts the "canonical" transcript, it depends on the context - which tissue you're looking at, for instance.

If you want to pick a generic transcript, I'd probably go for the longest of the coding isoforms and use that, although you'd find better meaning if you could handpick relevant isoforms for at least the important genes.