The final report to the funding agency, by Wagner, Betz and Koenig in 1990, conceded the general result: most of the several hundred participants did not perform well. What it went on to argue is that a small number had scored at rates chance could scarcely account for, and that this subset demonstrated a real effect in a few gifted individuals.
That subset carried the entire positive claim, and it is where the disagreement lives.
J. T. Enright reanalysed the same data and concluded that the positive finding rested on statistical treatment fitted to the results after they were in hand. His argument is the standard one against selecting a high-scoring group from a large sample: with several hundred participants, a handful will score well for the same reason a handful of coins in a large batch land heads several times running, and singling them out afterwards and testing them as though they had been nominated in advance turns noise into a result. Under conventional analysis, he reported, the effect was not there. Betz replied that it persists.
Neither side has been able to close the argument, and the reason is that the test that would close it has not been run. If the successful dowsers were genuinely gifted, they could be named in advance and tested again, and the effect should reappear. It has not been reproduced.
The atlas records the disagreement rather than a winner, because both readings of this dataset have been defended in print by people who worked on it.