A good question, JD33.
When I was a member of the Maths panel for a local CSE board we religiously determined the grade boundaries based on the rule that a median student for the year group as a whole (rather than for the actual number of CSE exam entrants) should be awarded a grade 4 (on an 1 to 6 pass scale). It wouldn't matter if every single candidate scored well over 90% that year; an 'average' candidate for the year group MUST only be awarded a grade 4, with only a small percentage (based upon a normal distribution curve) being awarded a (O-level equivalent) grade 1.
So, by definition, the CSE pass rate in Maths each year was always very similar to the one for the previous year. It could never be anything else!
However, at the same time, our colleagues on the English panel adopted the opposite approach, whereby they tried to define a certain standard of achievement applicable to each grade, irrespective of assumptions based upon normal distribution curves. So, surprise, surprise, CSE English results got better every year!
The GCSE panels seem to have adopted a criterion-based approach but with (in the early stages) the actual criteria being re-considered (i.e. lowered!) year upon year. In my own school days (irrespective of whether it was GCE, CSE or just an internal school examination) the assumption was that to achieve a 'top' grade in an exam you needed to score at least 90% (and, in many cases, as much as 95%), with anything much below 60% often being a 'fail' mark. These days it seems that in many exams (not just in schools), 70% will achieve a 'top' grade, with scores as low as 30% still being regarded as a 'pass'. In particular, that applies to quite a few university courses in Medicine, meaning that the 'fully-qualified' doctor who treats you might well have never score much over 30% in any of his exams!!!