The pilot must be capable of making a safe decision and skillfully use the flight visibility and visual references to visually control the airplane to land. The main problem, as I see it, is safely conducting the visual segment which means, to begin with, not accepting less than then necessary visual cues. The safety of every instrument approach, except Cat III autoland, requires this.
That is 100% correct and is the nub of the matter. And the decision must be that those "necessary visual cues" to continue below DH have ALREADY "been in view for sufficient time for the pilot to have completed an assessment of the aircraft position and rate of change of position, in relation to the desired flight path."
But we know that this assessment is very difficult in the vertical plane, and needs sight of at least a point of "low relative motion" (i.e. the aiming point). That is in no way guaranteed to be visible by DH, and is not even theoretically available in limiting conditions. Unfortunately the industry has for decades glossed over this inconvenient fact and trained pilots to believe that far less visual reference will be OK.
That means that the pilot's actual ability to "visually control the airplane to land" may have been fatally compromised without him even realising it. 99% of the time we get get away with it, but every now and again some unfortunate individual does not and is held responsible for making a "pilot error".