Per-capita values of countries with low population tend to be vulnerable to high variance
Yeah I was kind of trying to hint at that.
Luxembourg has a _lot_ of non-residents who commute there for work, and possibly drink coffee, which is inclined to mess up statistics like this.
They have tons of commuters and business travelers. Didn't look into the methodology, but my first guess would be that those are included in the numerator, but not in the denominator.