I am wondering if there is any need to align data at all if the goal is to quantify transcripts with salmon.
Are there any advantages to aligning versus doing direct. quantification using salmon, since both options seem to exist.
I have previously used HISAT2 mapping to ref genome and am wondering if I can perform Salmon quantification for transcripts on these BAM files.
No, if the goal is quantification using salmon, there is no need for mapping.
I think it would depend on your specie. Repeat content, G+C%, and multicopy gene families might affect the results, especially those produced by quasi-mappers (like salmon).
You can't use bam files as input for salmon, can you? Besides, salmon has its own algorithm for "mapping", using a bam file produced from a different aligner would definitely affect the result. From salmon's description:
Salmon can't use bam files, but it's easy enough to convert bam back to fastq files. I typically do this, but I don't have a fantastic reason why other than being stubborn.
If you don't have access to the fastq files yes, you can convert them, otherwise, I wouldn't say it is being stubborn.