It's better in the long run as you can compere results across cohorts fairly.
But cohort size changes (the number of babies born can differ substantially year over year), which impedes comparability.
Sure, both ends of the grading scale are easy to read when using bell curve grading. The problematic part is the center (where - by definition - the majority of graduates will be found): The C grade of year X could very well be better than the B grade of year Y.
Bell curve grading seems to be the worst of both worlds - the students experience even less correlation between effort and grade and you weaken inter-cohort comparability. If grading reform is willing to lose comparability between students anyway it should move in the opposite direction and grade relative performance improvements of individual students. That is proven to result in better overall performance of students (especially for weak performers).
The point is with scoring on a curve you can say that this person scored at this point in the curve so they are comparable to some one who achieved the same 20 years ago.
Intelligence pretty clearly has environmental influences and can change over time, even if the base value and/or maximum achievable level for a particular individual is genetically predetermined.
> The point is with scoring on a curve you can say that this person scored at this point in the curve so they are comparable to some one who achieved the same 20 years ago.
Comparable in terms of where they are in the distribution of people who were tested at the same time.
Not necessarily comparable in the actual skill that is being measured.
Which comparison is important depends on why you are comparing; there are times when you want the former kind of comparison, and times you want the latter. (If you keep both individual raw scores and group statistics, you can answer both questions appropriately, if you just grade on a curve and keep only the result of the curve, you cannot.)
Intelligence development certainly is both heavily dependent on environment and can change for a given individual if the environment changes.
A quite well designed study on this topic is the Minnesota Transracial Adoption Study that shows that adopted children have an IQ more similar to their social parents than their biological parents. It further showed that earlier adoption (= more exposure to new environment) meant higher correlation.
Whats more important however is the fact that IQ as measured is not a very good indicator for future performance. Motivation, especially achievement motivation has been shown to be at least as important for a successful professional life. And achievement motivation has been shown to improve strongly when grading based on individual performance improvement, e.g. measuring the relative performance gains (see e.g. Rheinberg & Krug, 1999).
But as it can be shown that the environment strongly influences IQ I'd say it is safe to assume that we as a species haven't yet approached the biological ceiling of our cognitive abilities. Once the environment doesn't influence IQ test results anymore, though...