The only thing I don't like about Pulsipher's comments are his stance that other ways of playing the game are inappropriate or not enjoyable.
To be fair to him, maybe other ways just hadn't been explored that much yet. Clearly the hobby has changed since then.
Extrapolation should be influenced by knowledge of the scenario. What do faction know and want, for example. What I don't want to do is make qualitative assessment of the players success and adjust the scenario on the fly either through adding or deleting elements of the situation or by adjusting original difficulty. "This has been a cakewalk; I better double the number of opponents!" "The PCs are too lucky; the BBEG shouldn't have failed his save in the first round I wanted him to get away!" "The PCs are really struggling; I think they'll find a new ally in the next room".
That's very difficult to do though if the players do anything that is at all unexpected.
For example, say a PC decides to try and rob a house. You were planning on an adventure elsewhere and didn't have any townspeople statted or know much about them or their protections or law enforcement. You might make one of several decisions; you might roll something to be noticed, you might let them waltz through, you might throw them a curve by having a significant NPC show up in some way. The point is, in an open-ended situation for which you aren't prepared, there's no unbiased way of going about resolving it. Any choice you might make is influenced by your knowledge of the players' actions.
Now, what I think varies will be the amount of improvisation a DM has to do. Like I pointed out earlier, if you have rigorous plans and the players stay within them, this conflict does not arise. I just think it's really difficult to consistently meet both of those criteria.
I have run different groups through the same scenarios in D&D and enjoyed the different play experiences and watching the different consequences unfold.
That's what I can acknowledge is possible to do, but is really unfathomable to me. Each game is a one-time thing; trying to run the same scenario again might cause my brain to explode.
And DM interference adds noise to the signal. If the DM interferes to adjust an outcome the players cannot use the result obtained as a fair data point. If the players do not know about the interference and do use the result as a data point, the model they develop will diverge from the game. In other words, they will assume similar interference as part of their model.
The model includes that interference though; it's called circumstance bonuses. (At least, some versions include that). And there's a real question of what you want the players to know.
If, for example, their attack bonus is +2 and the enemy's AC is 13, should it be knowable to the player that they have a 50% chance of succeeding at an attack? If they roll the AC exactly and then one point below it, they will conclusively know that. Conversely, if you occasionally throw in a circumstance modifier, they won't be able to reach that level of knowledge. I don't myself do this, but I could see it being done.
The problem is that if player knowledge exceeds character knowledge, you're metagaming. So if you're trying to avoid that, some degree of tomfoolery behind the scenes may be necessary to obfuscate the omniscient player's knowledge level down a bit in some cases.
You play out the chase to determine what resources are expended, if any and how much time was gained by the opponents, and if the PCs manage to discover the ruse and thus gain more knowledge of the opponent's abilities. In other words, you do it to determine consequence and situation extrapolation. So long as the figment was adjudicated correctly, of course. Most such devices don't have the capacity to travel far, act independently, or to respond to new environments.
I don't have any idea where it was or not; but there's a question in my mind here again about intent. If the NPC had some good reason for behaving this way, it's one thing. If the DM is metagaming this to get the outcome he wants (us finding an illusion), then it feels rather contrived.
I am willing to assume the author was trying to address DMs working at the table level as opposed to working with in-game motivations.
That may be; it's not clear what he meant.