In the second exam type, even a good student can't assume they can solve every problem within the time limit. They have to prioritize the problems based on how easy they expect them to be. The exam tests not only what the student has learned but also how well they understand what they can do and what they can't. The same can also apply to other types of assignments. For example, even a good student can't assume that they can solve every homework problem before the deadline.
The third grading approach is usually used for most important works. Under this scheme, the highest grades are reserved to those who do something exceptional. No matter what the expectations are, if you only meet them 100%, it's not sufficient for the highest grade.
As an example of the third approach, when I studied CS, our department gave the highest grade to only around 1 thesis in 100. There were no set criteria for the highest grade, except that the faculty had to agree that the work was beyond what one would expect from an excellent thesis. It was something not even a smart determined student could expect to gain, except by accident. Even the theses that got the highest grade were rather unexceptional in the grand scheme of things. Except the one that described an OS that runs on billions of devices today. That's the proper bar an undergraduate student should aim for if they want to guarantee the highest grade under this grading scheme.
This untrue implication. You can have difficult exercises that do require special insights with long solutions and still keep it within curriculum or "what you was expected to learn". What you cant do is to give them exercises that require one more chapter or invention of difficult to prove theorems that were not taught.
Meh, if your goal is for students to learn from mistakes, you can give them those any time during semester. No reason to put them on test. The best students will try to crack them, in fact, will spend more time that test allows on them. The difficult questions, especially in math or computer science, may take multiple days and attempts to solve. Going through that frustration is definitely learning experience. Putting time hour limit on it during test just weakens it.
> Highest grades are only available for those who demonstrate understanding well beyond the scope of the class or the degree.
In that point, it becomes in scope. Because "beyond scope" is infinite you have people guessing what "beyond scope" they should learn to gain it. The tests and exams should not be game.
Let it be know what are you going to test in advance.