How long was the review? It might just be that reducing the length/number of clauses helped. Honestly, if I was naively writing a restaurant review filter, length would be in there; if you look at TripAdvisor or similar, as reviews get longer the probability of the reviewer being a weird crank approaches one.
The length may have been a factor, but it definitely wasn't the only factor. When I eventually narrowed it down to those two paragraphs being the only difference, the attempt with the first paragraph in its original form and the attempt with that paragraph sanitized of profanity were basically the same length.
