True enough. But you also can't just wave your hand over them and cry "Numbers!" if they're not the right numbers for the question you're asking.
Look at your monkeys example. That's predicated on the monkeys coming up with a particular sentence. How do the odds change if we only ask the monkeys to come up with any valid English sentence? And how about if, every time they produce an English word, we let them keep it?
Have you heard of Richard Dawkins's METHINKS IT IS LIKE A WEASEL experiment? He ran the same test as your monkeys, but instead of forcing them to start over each time, he let them keep any letters they typed that matched the target sentence. It took them only 43 iterations to type the sentence.
Now, this isn't a perfect analogy to evolution, as Dawkins admits. For one thing, there's still a target, which as I've been pointing out isn't the case with evolution. And saving every letter isn't realistic; in reality, some letters would change back and then be found again on the way to the final sentence. On the other hand, nobody's claiming anything very drastic happens in only 43 generations. But somewhere between that and your example's umptymillion generations example lies the truth.
In other words, he rigged the test. Deciding what letter to keep and which one to throw away is bringing intelligence into the matter. You can't have it both ways. There is no preference to organization in the chemistry, hence the random trials. As soon as you make the test semi-random, you are appealing to some organizational vector that does not exist in nature as we know it.
And if allowing forms of life based on different chemistry makes it that much more likely you should see them on earth as well.
He actually deals with this issue in his follow on essay. The math gets a lot more complicated, but doesn't change the result that much. Any coherent sentence is still a very specific set of test results mathematically.
Actually... depending on the system, 43 generations is overkill. I think the longest I ever used a cell line in an experiment was 24 generations (about 12 weeks). Even with minimizing selective pressures, the cells simply evolve too quickly to be sure that the ending cell line is the same as the beginning cell line (and being able to show that is a must for getting the results published).
As for your overall discussion, the view that the literal creationists seem to be advancing now is that not only is it impossible for a nucleotide soup to spontaneously assemble into a human (or whatever endpoint of choice) genome, but that this spontaneous assembly of nucleotides to form a genome must occur at every generation in order for evolution to proceed. The premise is faulty from the get-go.