Can't get the paper link that you referenced to load right now, though I think I recall having seen this before. I'd offer a note of caution that rare events can play tricks on apparent signals if the sampling frequency is too short. Violent tornadoes are just so rare, and accurate F-scale ratings aside, small scale signals in the data might be looking too closely without a longer record. Might be there is something 'special' about these regions, but it could be something like a human factor, where perhaps a longtime surveyor at a local office tended to rate higher than surrounding offices. Just so many uncertainties in the storm data to look at it too literally imo.
Glen