The distribution of emulsion flaws was presumably random before the ones closest to the ecliptic were removed.
It wasn't. Villarroel's own data with regard to plate edges strongly suggest that the flaws/"transients" are more dense near the plate edges, though depending on the algorithm, this may be exacerbated by plate overlap. Remember, the graphs of the various sets show the plate grid visually.
We also have strong suggestions that emulsion processing was improved to reduce the flaws at some point. If that is the case, we can't even say that each plate ought to have approximately the same amount of flaws.
But if we assumed a random distribution of emulsion flaws, then, for a valid null hypothesis, the validation process needs to be applied to the random distribution. Since the validation process does not identify astronomical objects one by one, but rather eliminates all signals within 5 arcseconds from known astronomical objects, more flaws will be eliminated from plates that have more astronomical objects on them. Since the asteroids and the milky way are nor randomly distributed across the sky, this introduces a nonuniformity in the overall distribution that we have also observed in the graphs.
The hypothesis is that this nonuniform random distribution that results from this processing would also produce the shadow deficit that Villarroel observed. This seems all the more likely as the mathematical properties that would determine the altitude of the orbits have not been demonstrated in the data by Villarroel, and in fact some remarks in the paper led me to conclude they don't exist.
