feature request: output un-pseudoaligned reads

[rschulzUK]

WIth typical kallisto pseudoalignment rates of ~70% for human total ribo-depleted RNA libraries, an obvious question is what the remaining 30% of reads represent, especially when aligners like hisat2 report alignment rates >97%. Given the speed advantage of kallisto, a reasonable workflow would be to only align the un-pseudoaligned reads afterwards to see what they are and quantify them separately with something like featureCounts. A kallisto option for dumping the fastq records of un-pseudoaligned reads into a file therefore would be very helpful. Cheers, r

[arogozhnikov]

I'd suggest to output per-read alignment information if requested. This will address this issue, and also will help with comparing alignments