On some GARP Mocks it is a fifth observation. I know, we have discussed here it recently but this is exactly a very "noisy" examination, imo. To make things even worse, they usually put both solutions among given answers. I'm a bit sceptic that they would point both.
@Flashback It is due to the history: when we started training the FRM circa 2005, GARP's solution (to the problem: what is the 95.0% basic HS VaR of 100 observations?) was the 5th-worst because that's Jorion's answer. But the syllabus assignment for basic HS was (and is) Kevin Dowd, who explains why selecting the 6th-worst observation is a little better but also why either is okay; Dowd is mathematically stronger and explains exactly why the solution is ambiguous for discrete distributions. From my perspective, it's just a fact that the 0.050 quantile for an n = 100 discrete distribution is ambiguous. We carefully explained that to GARP and showed them why some of the extant questions were confusing to candidates who actually read the assignment (Dowd) on the subject. It took about three years to get it "sunk un," so to speak; this is typical of much of our feedback. They subsequently, I believe, revised some of the confusing questions, and the easiest way to revise was to edit questions so that either approach produces the same answer; e.g., a sequence of 100, 99, 98, 97, 96, 96, 95 ... would do the trick!). I hope that cures your skepticism