Hi @foreverwander, thanks again for your patience here!
So there are a couple of unusual things going on here:
The same barcode is being used for your forward and reverse primer - dual index barcodes are usually utilizing 2 different barcodes to increase multiplexing. When you de-multiplex with just your forward reads everything seems to work out, but when you add in the reverse barcodes is where we start seeing some issues. Something fishy might be going on with your barcode sequences in your reverse barcode file.
Regardless, if the same barcode is being used both on the forward and reverse reads, you could just use the forward only reads to de-multiplex since that has good results (which you have already done). Then in a separate step you can simply use cutadapt trim-paired to remove the primers from both reads which will also trim off the remaining barcodes/adapters.
It also appears as though the barcode sequences in your .fastq header say that your sequences have already been demultiplexed.
@Mehrbod_Estaki and @llenzi both expressed that they would like to see more than one line for the barcode mapping file, just to confirm that the same barcode was used in R1 and R2. This would be good, because q2-demux cannot be used in any other case.
Based on the fact that your barcodes are 12 bp in length, it appears as though they are golay barcodes - in which case, q2-demux emp-paired may be a better method in this case. Please take a look at this forum post for additional details:
@thermokarst was suggesting to add -p-rev-comp-barcodes as well as --p-rev-comp-mapping-barcodes in this case because the orientation is important for golay barcodes, which may explain the confusion on direction above.
These are all guesses here, so we may be off-track on this - but as a last resort I'd recommend taking these findings back to your sequencing provider to see if they can provide any additional insight on this. Hopefully this helps you with some next steps!
Cheers,
Liz