The linguistic quibble sources from Morris. Randi has an easy pedantic escape by claiming that there are no underground rivers. Despite what you say, they are paleochannels, or indeed, as you describe them, underground streams. Point is, I've absolutely no dount that Randi knows that there are many watercourses underground - but not as many as are claimed, and quite simply, there is no way that a dowser can pinppoint their location in the manner that they claim. His claim that there is water underground in the vast majority of spots drilled is quite correct too; that it isn't of use as a source is entirely irrelevant to the question. Indeed, the north-east of Australia, despite being largely desert, has a stable and enormous underground lake.
The claim, in yhem meantime, is no more than an irrelevance and an attempt to win a mill through pedantry - and through a level of pedantry that the claimant cannot survive when it is applied to his own words. As can be seen from the DNA example which he refuses to address.