Difference between the genome browser view of BAM produced by STAR and the result of Salmon

[NJU-Bio-Info]

Hi, such a fantastic RNA-seq analyze tool Salmon! Recently, I used Salmon to analyze transcript level expression and the information of my analyze tool:

• Salmon: v1.6.0 (install from bioconda)
• STAR: 2.7.9a (install from bioconda)

And the detailed commands:

• Salmon:
  
  and the transcript index is constructed with hg38 transcripts fasta download from GENCODE:
  

• STAR:
  

All commands were evaluated with Centos system.

After get the result file of Salmon, I tried to check the result in genome browser using BAM only with unique mapping reads produced by STAR. But I am confused with the following result:

According to salmon transript level TPM value, PPIE gene is tend to use longer isoform (the labelled value is TPM):

But when I visualized this gene in genome browser, I found that their are few reads aligned to specific region of the longest transcript:

So, I'm confused about this result and search form some possible reasons. There are some people saying that because Salmon uses transcript index rather than genome index like STAR, so it can rescue more reads. But I would like to ask you for more advices, great thanks!