First, thanks for the kind words. Now, I disagree with using the whole genome, due to the additions and deletions of genes and the fact that there are different time factors involved in the changes. So it seems to me a better indicator to select a fairly constant gene(actually the end product) and compare them. I selected the ADH3 protein from Ciona (AAL72131) and let Blast give me the answer. I used a 377aa sequence. The program evidently takes the weighted substitutions to give a score. How it arrives at the final number is a bit unclear since in the previous data Danio was given a higher score (1370) than Sparus (1369) even though in the following data Sparus has more (296) positive matches than Danio (295).
You can also see that the program searches for a match within each sequence because not all have 377 as a "denominator". Adjusting for that will not change the order(even though the lizard has one more positive than the mosquito it has four less identities.) You can subtract each positive from 377 to calculate your mismatch number
But in any case I'm wondering if you're using the program properly. If the three species have very different numbers of mismatches in the first 50 or so amino acids, the chances they end up with virtually the same number of mismatches over the whole sequence are rather low.
I'll run the sequences by hand, and get back to you. I don't trust canned programs unless I've checked them by hand first.