Your calculation has no connection to the convergent evolution described in the papers you cite. You're calculating (incorrectly) the probability of identical genes being formed by chance in multiple lineages, starting with completely unrelated sequence. What's described in the papers are changes to the different versions of the same gene in different lineages, such that the copies become more similar (in amino acid sequence) than expected by chance. The actual number of changes is small. For example, the researchers looked at three echo-locating lineages for the gene Cdh23, one mutation occurred in all three lineages, an additional 20 mutations were shared by one pair of lineages, and an additional 3 mutations were shared by another pair of lineages. That's not remotely like the situation you modeled.