Yes. That’s the very thing I am saying is opposite to what you are claiming.
There is no rhyme or reason to the pattern of WNT genes in invertebrates. It does not follow the tree at all.
There’s lots of rhyme and reason, you’re just not allowing yourself to see it. It is certainly noisy, no doubt about that, but there is nesting hierarchical structure even in that small data set.
Look at the phylogeny and see what is most closely related to what, according to the tree shown on the left. Homo sapiens is most closely related to chicken (Gallus gallus) of the species shown on that tree. Is that fact reflected in shared and missing genes between them? Yes, between humans and chicken, we need invoke no losses. WntA is absent in both, so that is probably an inherited ancestral state. Moving on, next most closely related is Xenopus tropicales (some frog species iirc). That’s a vertebrate too. Is that relationship also reflected in the 13 loci used? Yes. Again WntA is absent, so it’s probably an inherited ancestral state. Again no losses need be invoked. And so on and so forth.
In general we need to invoke fewer gene losses for members within a clade, than we do between different clades. And the gene-losses generally follow the nesting hierarchical pattern. The genes lost in some species within a clade, generally are more likely to predict other genes lost by other members of that same clade.
Even though there’s basically only 13 sites used, and they’re only counted by binary presence or absence, there’s still obvious nesting hierarchical structure in that data set.
But I’d like for you to understand and spend some time thinking about how simple and small the data set is.
At least with DNA sequences, you still have 4 possible states for each locus. Here we just have two, either the gene is present or it is not. And we only have 13 such sites used. It would be like trying to generate a phylogeny based in a nucleotide sequence of about 6-8 bases.
Still, we can see the pattern even here, despite how little data we have to work with.