Paul Newton, of the Qualifications and Curriculum Authority, has called for a more "rational" approach to setting standards, with boards saying explicitly how they ensure, for example, that all A-levels are the same standard.
If they had been more open about these arrangements, he argues, then the A-level re-marking crisis of 2002, which led to two official inquiries into last-minute changes to students' grades, might have been avoided.
In an article in the journal Assessment in Education, Mr Newton said that the QCA's code of practice, introduced after the 2002 fiasco, still leaves areas of uncertainty for markers.
Examiners use both statistical evidence about the performance of students and their own judgements of the quality of work to decide where to place grade boundaries.
But Mr Newton said the code gives no guidance on how boards should react when statistical and subjective evidence conflicts, potentially creating problems for the heads of boards.
He then analysed arguments put forward in a previous research paper by Mike Cresswell, now director general of the AQA board, Britain's largest, in defence of the current system.
Dr Cresswell essentially argued that ensuring consistent exam standards between subjects and over time rests on value judgements by examiners, and that it is not possible to lay down precise rules by which these judgements are made.
Mr Newton said this meant the system was hard for the public to understand and potentially unfair to candidates who were not told the rules by which they were judged. When it is not possible to say standards between subjects or across years are the same, this should be made clear.
He said boards had claimed since the A-level's introduction in 1951, that standards had been preserved from year to year - yet no-one would argue that the pass standard 54 years ago was the same as it is now.
Confusion over maintaining standards between old-style A-levels assessed before 2002, and the current qualification, lay behind the regrading fiasco, he said.
Dr Cresswell could not be reached for comment.