First absurdum:
100 elementary school students take HKALE Pure Math examinations. Betty get 1 mark out of 100 while all other candidates get 0 marks. As a result, Betty gets an A grade.
Second absurdum:
100 postgraduate Math lads take HKAT exam (i.e. primary six level). Johnny gets 99 marks out of 100 while all other guys gain full marks. As a result, Johhny gets an F grade.
The above absurda show that norm-referenced grading is pretty normal distribution. Such a grading system only mirrors how candidates are ranked, while the real strength of every candidate is neglected. That's why HKDSE uses solely standard-referenced grading.
By central limit theorem, even though HKDSE uses standard-referenced grading, when the sample size is large enough, the final grading will still observe normal-distrbution. This mirrors a fact that the standard is not absolute enough.
So, I would like to recommend dual grading system, i.e. absolute standard referenced grading alongside the conventional norm-reference grading. The absolute standard would be set as below:
<25% of total subject mark: UNCL
<40%: Level 1
<50%: Level 2
<60%: Level 3
<70%: Level 4
<80%: Level 5
<90%: Level 5*
<100%: Level 5**
To engineering school benchmarks, scoring 90% of total marks is the minimum threshold of graduation. So you see, current HKDSE grading has a fatal loophole.