Variations of incidence/prevalence of disorders with the season of birth make a lot of sense. Schizophrenics are more likely born in winter, for instance. The differences are tiny and only found in hugh samples.

Noone (among those doing this research) believes in sun signs or months as the real causative variable responsible for the differences. The real variable(s) confounded with season of birth could be:
(1) food intake of mother during pregnancy (varying with season, of course)
(2) weather during the first months after birth or during the months when the kids starts walking and crawling (different early experiences)
(3) social background of the kid (the different social strata have slightly different maxima of their kids' birthmonths)
...

But the idea of the study above was different: gather a lot of data and look for a lot of differences, then you are sure to find some statistically significant differences by chance alone. That's the unreliable method of "data snooping". Good for gathering hypotheses, bad for testing hypotheses. The remedy is the split half method: take half of your data for gathering hypotheses and the other half for testing them. If something then remains statistically significant, pay attention to it, if not, forget it.

The above study has deliberately made that one mistake for the purpose of demonstrating that error and later shown what the remedy is.

I am very grateful to know about it, for I can use it well for demonstration purposes.

Wolfgang