Of course it does, in a population of 1000 and using the assumptions I offered, it cannot be any other way.<quoted text>
By your analysis, in a stable population, virtually every mtDNA haplotype introduced disappears in about 24 generations. To you does that seem to pass the sniff test?
Of course, you also have to remember that new novel variations are being introduced all through these generations as well. To get a clearer picture, now imagine that a single new variation is introduced at random among one of the 1000 women in each generation. So as current numbers are dwindling, new variants are emerging. That should make it interesting. And don't forget, that because the new variant strikes at random, it can either hit one of the original haplotypes or it can hit one that already underwent a variation previously...you could call that a "granddaughter" variant of the original type.
You can see the development of a nested hierarchy in action here.
You can also begin to see why a timescale of 5000 years is not possible no matter how you try to stretch or squeeze the assumptions.