That's why there were experiments with different wording of the problems. All of the logical setups and solutions are identical (when expressed in abstract P/Q, true/false), but some problem domains (took a plane) is a lot more difficult for average people to comprehend than others (ate the vegetables).

This is not in the article (I'm just familiar with the exercise), but people are also much better at dealing with logic when the questions are about legal drinking age.