Agreed - it would be interesting to run the numbers with "questions may not have an answer" as a qualifier. If I try to put myself back in my young student mindset, I would only expect to be correct with a non-answer response for a specific kind of test, or with a certain type of teacher that asked non-structured problems. If given this problem in a structured environment such as a standardized test I would likely look for a possible answer, realize none was possible, write a number as a guess with some notes about the median age of shepherds, and then be extremely frustrated at the test designer haha.

I get that this is supposed to outline the differences between structured / unstructured learning, thinking, and classroom conditioning, but it's not quite fair to draw a conclusion that doesn't take that conditioning into account.