BB, A very interesting view.
A difficulty in explaining performance is that most activities involve tacit skills; as learnt with experience, but hard to describe. You can read or be told how to ride a bike, but you still have 'to do it', learning though practice.
The standard of human performance depends on context.
In normal operations we should strive for the best, but reality and human behaviour often accepts lower standards - we 'satisfice' according to the perceived situation and level of skill - we are lazy, minimising cognitive resource. In these circumstances modern automation is able to exceed human capability.
In abnormal situations, no automation, then the human will always be 'best'. However, the level of performance need not be as required for normal operations, just match the conditions and minimising the risk of harm - redefined objective.
Observing and participating in research into human capability in high workload - novel situations - manual landings in fog; the human is more capable that imagined. An external view might quantify this, but internally this cannot be judged until experienced, thus generating a concern about ability - confidence.
Limiting conditions were very difficult to define, best identified by outcome 'I didn't expect that', but that's when we learn.
Perhaps normal operations do not allow - accept sufficient 'I didn't expect that' situations from which to learn; yet with rare failures in modern operations it doesn't matter except for dented pride - is that what we fear most of all.