what might be going on to have so many bad index reads in skim-seq data

Question

We have pair-end reads from skim-seq (Nextera) data where we found only 60% good barcodes (i5 and i7). When we demultiplexed them we found the read count in chromosomes (per