The general assumption is that if you're looking for a transcript that falls below the detection threshold of RNAseq, then it's likely to be so weakly expressed as to be biologically negligible. Furthermore, with the bursty nature of gene expression, an average of 2-3 copies per cell (about my limit of detection in a reasonably sized experiment) could still have a sizable fraction of cells with none at all.

That said, the downside of the Oxford Nanopore for RNA-seq is that, while you get longer reads, it's not yet really clear how many reads you get, which is at least as important for trying to find those moderate-lowly expressed genes.