As often happens in these situations, I'd suspect that a substantial part of the tension is misunderstanding about the meanings of the "expectations". (That there'd be misunderstandings is not surprising, considering that the prof and the student are in very different places in their experiences and their lives, and almost surely attach significantly different connotations to words...)
On one hand, _of_course_ it makes no sense to agree to, as they say, "a pig in a poke", that is, something unknown. Perhaps to gamble, etc., but...
The slightly different remark/question I make to my prospective students is more about "trust", than "compliance". I do try, and do try to explain to students, that the projects I suggest to them are not only potentially fruitful, but tailored if possible to my observations of their tastes and talents.
(In math, in the U.S., at an R1), if I were somehow required to fully justify to a skeptical student the sense/competence of my recommendations to them, I think we'd get nowhere. Certainly in contemporary number theory and automorphic forms and such, there's a huge backstory to be assimilated, and a huge technical "library" to appreciate, before most things are truly intelligible. I cannot impart this to a novice in a conversation.
So, although I do appreciate the genuine intellectual appropriateness of a wish to have things explained (rather than accepting things "on faith"), the situation of "getting a PhD in six years or less" seems to require a significant element of trust...
As in many other human situations.
EDIT: in response to comment... in math, in the U.S., there is in general no immediate research benefit to "having Ph.D. students", or postdocs. That is, there is scant "grunt work" to be done. So taking on PhD students or postdocs is giving something... which, in the style of authorship of math in the U.S., will not result in any authorship at all.
So, when/if I take on a student, I have no anticipation of any substantial administrative reward or return. It's because I like helping people learn how to do (to my perception) fairly amazing, cool things. If a student disagrees with me about what is amazing or cool, that's fine, but obvs I can't be an effective advisor.
Also, sometimes... and to my mind these are some of the best times (in math), I may have some vague intuition that some line of inquiry would be good. Sure, a student does not have to believe me. But, on the other hand, they should not have me as their advisor if they don't trust my "intuition" based on decades of experience...