We all agree that it is difficult to define the knowledge of a foreign language, very difficult to evaluate the oral knowledge of a speaker, and almost outrageous to claim to give a score for that.
The fact is, that there is a difference among someone who is barely able to present himself, someone who can have an everyday conversation in a restaurant and someone who can deliver a speech to an university audience.
And if you want to offer exams, you must have a transparent way to say it, to measure it, and given that the measurers are human beings, the way has to be the most independent and less biased possible.
The actual system of scoring in the DALI examination is already a good basis (see previous post), one way to make it better is to increase the grades of evaluation; where you have now four levels, can split them to arrive at six, like this:

OLD
statement formally correct or with some small grammatical mistakes (2 points)
statement mainly correct, with grammar mistakes which do not compromise the understanding (1.5 points)
statement mainly incorrect, and partially not understandable (1 point)
statement not understandable, or no statement. (0 point)

NEW
statement formally correct (2.5 points)
statement with some small grammatical mistakes (2 points)
statement mainly correct, with grammar mistakes which do not compromise the understanding (1.5 points)
statement mainly incorrect but generally understandable (1 point)
statement partially not understandable (0.5points)
statement not understandable, or no statement. (0 point)

This way, even if the experts continue to disagree as before, the wider score should “buffer” the differences and the final evaluation should result more homogeneous.
(end of part two of three)