You didn't explain carefully what your data consists of, so it is difficult to help. From what I gather, you have an (unknown origin) assembly, and want to polish it with the raw PacBio sequencing reads.
Are you trying to merge the
.bam.pbi files, or the
.bam files? The
.bam.pbi files are PacBio BAM index files. In addition to samtools not working on them, what you want to merge are the
.bam files, then create a new index for the merged bam with the
pbindex program. The BAM recipes wiki has useful information regarding handling of PacBio bam files.
I don't have experience with Arrow and other PacBio tools, but it seems you have to use an aligned (to the assembly you want to polish) bam with Arrow, not the original unaligned bams you have been given by the sequencing center. The docs I've read used BLASR for the alignment step, I don't know if aligning with minimap2 would work with Arrow.
I was planning to index, align, and sort after merging before running arrow.
In case you didn't merge the unaligned bams, you can first map each subread bam separately, sort each of them, then merge the bam, which will result in a sorted merged bam. After merging, you can index the bam, and use this bam with Arrow.