Boci has it right with this:
And the added context is that there is an argument that having more smaller attacks is better because they can minimize damage overflow, even if the overall damage dealt is smaller. There is a theoretical space where this is the case. Let's say there was no random rolls for hp or damage. All the enemies have 8 hp. Player A gets 2 attacks that do 7 each, for a total of 14. Player B gets 3 attacks for 4 each, for a total of 12. It takes each player 2 hits to drop an opponent. Against 3 opponents, Player A takes 3 rounds to do this, while Player B takes 2 rounds, even though his damage output is lower. The problem is that, in a lot of 5e, it is more like player A dealing 7 pts. 2x a round and Player B dealing 4 pts 3x a round against 4 opponents, each with 50 hp (those darn bags o' hp everyone keeps complaining about). So taking 8 strikes at 2/rnd or 13 strikes at 3/rnd, well it looks like the overall average damage becomes a greater factor than any loss through damage overflow.
So Knaight's point isn't wrong on the level of "things other than statistical average DPR are important to a real-gaming analysis." That's correct and he's good to have seen beyond the white-room analysis to recognize it. However, the setup of the game, as it works, based on the opponents and how they are constructed, tends to work against the advantage of the adaptability he is espousing as the benefit that redeems dual-wielding.