What I've come to realize is that a mechanic based on luck and a mechanic based on skill both serve the same purpose in this final framework. They both act as a "scattering" agent: they push the player in somewhat unexpected vectors, allowing them to explore the gameplay "terrain" with a quick and interesting feedback loop.
Generally, I agree with him, but his language is abstract enough that I think it could really do with some more concrete examples. Let's consider three turn-based games with units on a grid: Chess, Civilization, and Archon. In each case, if one unit attempts to move into a space occupied by an enemy unit, this creates a conflict, which needs to be resolved somehow. In Chess, a consistent rule is applied - the piece that's moving wins the encounter every time. In Civilization games, resolution depends on a combination of the units' stats and random luck (if the units are of similar strength, considering any modifiers, you've got about a 50/50 chance of winning), and in Archon when two pieces overlap it causes a dueling mini-game to start, where the pieces, controlled by the two opposing players, duke it out. The point is that in each case the overall strategy of moving pieces on the game board will be similar, but the difference between Chess and the other two, in terms of strategy, is that you can't afford to be quite as precise and pre-planned in Civilization or Archon because an enemy unit could unexpectedly break through your front line if your opponent is lucky or good at the minigame, respectively.
Avid players of pen-and-paper games might be more willing to accept the interchangeability of twitch-challenges and random chance since their games so often rely on random elements to resolve situations that theoretically rely on the character's skill. A d20 roll in D&D could be replaced by rock, paper, scissors (which is still mostly luck-based), or even by a quick game of table-tennis while retaining the overall structure of the role-playing game. Video gamers may be less likely to accept this interchangeability because they're more likely to experience the twitch-skill end of the spectrum and value the skills they've developed. But from the game designer's perspective, the range of skill-levels of the players represents a predictable curve of possible outcomes, just like the die-roll does.
This isn't to denigrate twitch-skills as being no better than luck, but simply to suggest that in any game where a the result of tactical decision isn't guaranteed, the designer can choose to add a twitch-skill element or an element of luck, and expect it to produce a similar "scattering" effect on the results. Generally speaking, the advantage of using skill is that it gives the player a greater sense of ownership over their victory, but it also makes it possible for a player skilled at the twitch-challenge to skew the results of tactical decisions more consistently then they could with a luck-challenge, whereas the luck-challenge preserves, over time, the relationship between strategy and expected success rate. Allowing a twitch-skilled player to surpass expectations may be fine in a single player game (I was really good at the lock-picking mini-game in Oblivion), because it's only that single player who's fun we need to worry about in the first place, but in a cooperative game such as D&D, a single player outshining everyone else in his party as a result of being a good table-tennis player could produce ill will, and I imagine this is part of the reason twitch-challenges are more likely to appear in video games than tabletop games (that, and the fact that twitch-challenges are easy to introduce in the medium of video games).