I've been exploring some STAR-aligned tumor RNA-Seq data in IGV, and noticed that coverage for either of the two reference coding transcripts of one particular gene (per Ensembl) is low throughout (e.g. some exons with 2 reads, some with 0) except for ~100 depth for one exon at the end of the gene.

Any thoughts on how I should interpret and further analyze this? Is this likely to be due to bias in the sequencing or is it suggesting that only that particular exon is expressed?

Relatedly, when I run Kallisto using --genomebam and view the pseudoalignments in IGV, I see a similar distribution except the depths below these exons are multiplied by an order of magnitude (thousands instead of ~100 reads overlapping that final exon). I'm trying to figure out that order of magnitude difference.

I'd appreciate any insight into either of the above questions. Thanks so much!