I have spent many happy hours working on marine electrical systems soaked in water / seawater / bilge etc. When thoroughly wet, I can tell whether a 12V DC circuit is energised by putting my finger on it. Grasping a 24V terminal with wet hands while sitting on a steel deck in wet clothing can be quite unpleasant. I would not be surprised if they can feel 18V.
What is important here is that the measurement of voltage, in itself, is not terribly informative because the effective source impedance is clearly quite high and the instrument input impedance will likely have a strong effect on the voltage indicated. What is relevant is that the voltage is not negligible, and that ought to be sufficient to make it traceable.
Have you had a chance to take readings, under the same conditions with water flowing, of the following set of three voltages:
a) External test spike to MET
b) External test spike to energised water stream
c) MET to energised water stream
If the conditions are identical and the spike outside the resistance area of the fault, then it should prove conclusively which point is being elevated above true earth.
Also, do you have an IR value for (suspect lighting circuit L+N) to (running water)?