Salmon Discarding A Read in Pair

[ericbretz]

Hello, I've run in to a situation that I can not wrap my head around and was hoping for some insight.
I've simulated reads for this single contig assembly using art_illumina. It is completely error free and has 100% coverage with 60,000 reads, 30,000 pairs.

When aligning with hisat2, the output SAM is as I would expect it. 60,000 pairs perfectly aligned.
However, when I try to quantify the output from hisat2 through salmon, exactly one half of reads are discarded. Every pair loses one of its reads resulting in 30,000 unpaired reads.

salmon quant --libType A --alignments single_aligned.sam --targets single.fa --sampleOut --output ./

I don't understand why this is happening. If there's a safe way to upload the .fa/.fq I would do that.

Example Reads

Geneious