The first problem with risk assessment matrices is awareness. You can't quantify what you don't know.
The next problem is that issues don't always have the same 'weight'. In some circumstances X might be inconvenient, in other circumstances it could be fatal. Giving X a value of 4 creates the illusion that the weight of 4 is definitive in all scenarios.
Risk matrices are necessarily limited to the master list of statistical issues. If the issue you anticipate is not listed, does that mean it doesn't exist? The RAM I must use has a final block for other in which I can enter any value deemed appropriate. Which is brilliant except that a pattern has been formed of values of 1-5 in the previous blocks, so the implied maximum is 5? The issue forms part of an acceptable risk and I should accept the request?
The fundamental ability required of the PIC is the ability to continuously evaluate conditions, ability, and plan survivable alternatives. A signed form with 23 points or 'low risk' or 'risk alpha' doesn't do that, neither does exercising your mathematical ability with the RAM score as the flight progresses. The 'plan' was based on what one believed the conditions might be in execution, and unfortunately, like Schrödinger's cat, everything changes when it becomes actuality. What tool do we have to address the fixed idea that flight we're on isn't the one we planned? That's what causes accidents.