Let us modify the example to show this point. Suppose the owner of the remaining water hole had made efforts to prevent it from drying up. This effort changes the whole story and Lockean proviso no longer applies. What is the difference? If the diligent owner didn't exist, this water hole would have dried up as other water holes, and the situation of the people would not be better than the appropriation of the water hole. In other words, the situation under appropriation is not worse than the "base line." This base line depends on the history, so the Lockean proviso is not an end-state principle.
It seems to me that there is a conceptual confusion in Nozick's argument; that is, the confusion between the existence of appropriation and the existence of the person who appropriate. In the modified example above, we can distinguish three possible situations.
(a) the owner doesn't exist (thus, the water hole has dried up).
(b) the owner exists, and appropriates the water hole.
(c) the owner exists, but doesn't appropriate the water hole.
Among these three alternatives, it is true that (b) is better than (a) for everyone, but still (c) is better than (b) for other people. Since we are talking about justification of the appropriation, it seems natural to compare (b) and (c). His introduction of (a) seems rather ad hoc interpretation of Lockean proviso to avoid an end-state principle.
Another problem comes from the fact that (a) is a counterfactual condition. How do we know what would happen if he didn't exist? Someone else might take care of the water hole, or his existence itself might have weird causal relations to the drying up of other water holes. Or his efforts might have nothing to do with the preservation of the water hole (can a rain dancer appropriate rain if people believe his dance caused the long awaited rain?) . With these essential uncertainties, Nozick's baseline become somewhat arbitrary. If we use (c) as the base line, we get much more straightforward answers. Here again, he need to explain why he use (a) instead of (c) as a base line with all uncertainties.
I think that the Lockean proviso should be used between (b) and (c). Of course this is not favorable for Nozick, since this version of the proviso is an end-state principle. But if he wants to avoid this conclusion, he should show how he can interpret Lockean proviso in his way without ad hocness, and how he can apply it despite the uncertainty described above.