Agreed, it simply looks like a missing step talking about variant filtering.

Note since this workflow was written, we also added samtools consensus which may or may not be a better fit. It's a totally different beast as the purpose of this is to produce a (potentially heterozygous) consensus given a BAM file and is aimed more at refinining assembly consensus post realignment, as a means to generate a reference-synchronised fasta for CRAM embedded references, or as a fast alternative to things like ivar consensus. However that said clearly there are similarities between the two pipelines. The samtools method is far simpler, faster, doesn't rely on first having the reference to start from (which someone knew in order to do the original alignments, but this may be not the person with the BAM/CRAM), and doesn't have the frankly bewildering calling conventions of bcftools consensus. However it's also much more rigid and lacks the huge flexibility of bcftools or the ability to do filtering first. It's also considerably less robust around indels as it has no indel-realignment step, so be wary of that around long-read technologies.

Anyway, I mention it as it may be a viable alternative to some users.