Yes, I agree genes counted from best matches to human, etc. using blast
are not the best way to analyze phylogeny (this is a "factoid",
implying you should take with some grains of salt). However, these
are for all the genes of these genomes, giving it a bit more strength
from numbers, than for instance if it were for one gene family.
So 600 Drosophila mel genes score as best matches to 15,000 human genes,
Tribulium has 2,000 best, and 3,300 Daphnia genes are best matches.
These are large differences. Using bitscore as a shorthand for quality of
alignments, there are larger differences. I haven't the time just now
to run this thru a more careful alignment and phylogenetic assessment, but
I'll wager these differences remain. Some others folks have
done such, but perhaps not with a recent set of arthropod genes.