Why must the Person–Action–Object order always be fixed?

Because you decode by slot: the person tells you the first item, the action the second, the object the third — swap the order and the scene becomes ambiguous.

A composite scene only reverses cleanly if you always read it the same way. If "Einstein (person) juggling (action) a violin (object)" always means item-A · item-B · item-C in that order, you can recover all three. If you sometimes put the object first, you can no longer tell which item each slot came from. Consistency of roles and order is what makes PAO decodable, not just memorable.