If students are going along solving word problems that always make sense, they have zero practice in identifying which questions are well posed and which aren't. It may not be that they're bad at reasoning through word problems, so much as they aren't used to word problems that make zero sense.
Given that context, it makes partial sense for a child to think "I must be making a mistake--I can't see how this problem works, but it must work somehow, because these problems always do make sense, even if sometimes I get them wrong, because I am not smart enough to understand them."
Is that a good attitude? Heck no. If nothing else, this experiment shows that we need to teach students that math problems aren't always going to be nicely set up so that they make sense. But that's different from thinking kids are dumb for not adapting to an unexpected wrinkle on a test.
An interesting companion experiment would be to prompt the students to explain the problem first, or see whether they could accurately determine which problems made sense and which didn't, when they knew that was what they were doing. My prediction is that the results would not be great, but that they'd be better.