First, there aren't three things, but two: bioseq and bioseq.pod.

Second, that metacpan lists this three times in two lines with one name duplicated in https://metacpan.org/source/ROCKY/Bio-BPWrapper-1.02_02/bin is its doing. And the second entry is backwards; the name listed is actually the documentation and the documentation associated with that name is really the command-line program. And that is a bug.

As for pod tools "knowing", if that's the thing that is deciding this, well it somehow knows that for a perl module. If there is a file that has extension .pod with the same name as he Perl Module, then that's not a separate program but documentation for the corresponding perl module.

This is what LanX's answer in Using .pod as a standalone file rather than in .pm and it showing up in MetaCPAN is about. I am sorry I didn't link that before.

Finally, what Module::Build may have to do with this is possibly how this stuff gets tagged or set up in a tarball: the pod in the bin directory is installed same as without the pod. It is possible and likely this is that's not dictated by Module::Build, but the tools and rules for processing a tarball: what and where to install. In this case it happens to have been created by Module::Build so that is one interface for controlling the building of the tarball. Sure, there are other things too that influence this as well.

Here is idea, instead of bin/bioseq.pod, save it as lib/bioseq.pod

Good suggestion. I tried that but no dice. The lib file looks nice but it is not associated with the bin.

At this point, I think the best alternative is to automate pulling in the wiki POD and stuff it into the command-line script after an __END__ just before doing a CPAN release.

but two: bioseq and bioseq.pod

Maybe if the script were named bioseq.pl it would work.

Maybe. But I'd still have the problem that the pod would be placed in bin which I don't want.