Has anyone looked at Paul Hsieh's SuperfastHash http://www.azillionmonkeys.com/qed/hash.html or Bob Jenkins even later and faster lookup3 both of which claim to be both fast and very evenly distributed.

I tested these as well as a large number of other hash functions. I even downloaded and compiled SMHasher and performed tests. When testing the hash tables in mIRC using direct calls to the hash table routines, FNV1-A-mod was better in both distribution and speed than any of the hashes tested so far. It has its flaws, eg. zero sensitivity, but that is unlikely to be an issue in practice. It's also satisfyingly simple :-) That is why I have chosen it.