Whether a program or a proof is "correct" is fairly objective. But there's a couple places where subjectivity enters in.

• Suppose you have an incorrect program/proof. How much partial credit does it deserve? How bad is one mistake versus another?
• Suppose the student's answer is correct but ugly. Classes routinely factor "good programming style" into the grade.
• Some work requires written answers or explanations; these can be good or bad.
• Upper level courses often require students to present a topic to the class. (Often by guiding the class through an important research paper.) Quality of presentation is graded subjectively.

In practice, there are important mitigating factors. Big lower-division or required upper-division undergraduate courses are autograded as much as possible, reducing subjectivity. Elective upper-level and graduate courses tend to give As to everybody anyway, since the professors want to keep people in the class and don't want to make trouble for "their" students.

(In American universities, often there's a rough division between first-two-years (lower division) and last-two-years (upper division). Upper division classes are normally for students who've already been admitted to the major, lower division will include prospective majors as well as interested outsiders.)

There are plenty of issues to disagree over. I remember some argument over what issues were important in program efficiency. He was probably right about that. I was dismissive of the practicality of pure LISP with no extralogicals and no sequencing (no 'seq' or indexed iteration). I was probably right about that.

It didn't help that I was a bit of an arrogant twit at the time.

But, the key wasn't grading. The most important factor was his claiming, at the end of the semester, not to have received an important homework from me. I had thrown it out by then, so I couldn't prove I'd done it; and he gave me a zero on it.

This could have been accidental. But it never happened to me in any other class.