EF4 or EF5 tornadoes are around 2% of those in the US (https://en.wikipedia.org/wiki/Tornadoes_in_the_United_States), which means any state with more than 34 or so a year can expect to see at least one. That accounts for 14 states with a combined population of about 106 million people. That's approximately 32% of the total US population.
8%-ish of tornadoes rate EF3. At which point any state with seven or more a year should expect at least one in the EF3-5 range (yay, birthday paradox!). That's 32 states totaling 270 million people - 82% of the US.
I find a flaw in your logic, but don't have a pithy name for it. Population densities and EF4/EF5 tornado probabilities are not uniformly distributed throughout a state, and probably are negatively correlated.
Looking at this image (https://en.wikipedia.org/wiki/File:Improved_Average_Annual_T...) from the wikipedia article you linked shows this, with New Orleans, Atlanta and Miami being outside their state's main tornado zone and Houston and Chicago right at the border of high-tornados and medium-tornados.
Oh, you're absolutely right. I simplified immensely by aggregating at the level of a convenient political division that has only vague bearing on actual geography as well as assuming the distributions are uniform across the range in question.
I think the overall point is sound, though. A huge chunk of the US population is in territory regularly subject to strong tornadoes.
Also, I would beg to differ that this map puts Chicago at the edge of a high-tornado zone. It's solidly in one, as is Kansas City.
