Am I missing the point? It wouldn't be the first time, but I don't think so. At the very least, it must be a far looser definition of "work" that you are using. If by work you mean simply "people who dowse can find water" then yes, they can. But then, that was never really in doubt. On the other hand, if by work you mean "produce results that are better than would be expected through random chance", then no, dowsing does not work.
Put another way, as far as I see it, by your definition, if dowsing works then so too do the Native American rain dances. They danced, it rained sometimes -- not consistently, of course not, no-one is claiming that -- but therefore it works.
There is no difference. You are saying that dowsing works because the intended effect follows sometimes. I am saying that it does not work because, while the intended effect did follow, there is no evidence of any causal connection between the method and the results.
Earlier svejk claimed that he could achieve a phenomenal 100% success rate, so that it has been suggested that dowsing works consistently.