Calibrating:
After much testing, I've resolved that the margin for error in sighting horizontal level through this device is 1/8" of camera height adjustment, within which I cannot distinguish high or low based on leveling of the pair of horizontal sighting strings.
My planned camera set back had been 36" but that produces a +/- margin for error 17.25 miles away of +/-162', which is nearly half of my observation elevation and that of the Central Coronados island. Even though I'm not trying to take sightings to verify the height of the island's summit, it illustrates the sensitivity Tom was talking about with these kinds of observations.
Moving my camera setback to 60" will make the margin of error more comfortable, especially since I'm not measuring the dip angle but just assessing -- pass/fail -- if the horizon is at eye-level. This will just add some degree of assurance that the distant island's summit is, in fact, in transit with 400' level of the eye from the observation point.
I think I've done all the prep and planning. Gotten some preliminary photos. Now it's just a matter of watching to see if that horizon line ever moves upward , using those islands as an index (with my "perspective cube" as an aid.
If that island summit is eye level, and the horizon never rises to it, then the test verifies that the claim "the horizon is always at eye level" fails.
If the horizon does move, to within at least half of the island's height (allowing for about 194' margin of error), then the test verifies that the claim "the horizon is always at eye level" passes.