This is missing the biggest piece: phylogenetic analysis. They aligned a selected group of mutations and then eyeballed the alignments and then speculated.

Here’s what the methods section for this paper should look like in order to make the theoretical leap.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8969601/

A reasonable phylogenetic tree is here: https://nextstrain.org/ncov/gisaid/global/6m