A pretty silly study on the whole; the issues of small sample size and high variance have already been pointed out by PBR. Even Horn spots the methodological issue of not using a known-nonexistent deity or similar as a control (to which I'd add, they should also have considered using one from a culture foreign to the study participants).
A possibly more serious issue is the issue of conflation of "believe" and "have an emotional reaction to", the problem with which should be obvious.
Horn also demonstrates his ignorance of the nature of delusions (quoting the idiotic "poached egg" comparison). This is best answered with this piece from Scott Alexander.