Yeah, something is still wrong up here. After flipping the barcodes, the 8 samples that demultiplexed all have acacacac in the seq_barcode_F column. (or perhaps in the seq_barcode_R column if that's how you flipped it)
This gives me less confidence that either way is correct 
Was the sequencing center able to successfully demultiplex your samples? It's possible we are still missing something, but if they are using a custom protocol, they may have a way to solve this problem. If so, they could send you pre-split results with each sample in a pair of fastq files, and we could take it from there!