What sticks out to me is that he chastises the old d100 NPC traits tables as being uninspired, as he sees them as "Just enough to distinguish him from the next dwarf, but nothing that's actually particularly meaningful or interesting", yet he goes on to say that what he looks for most boils down to what is essentially just a collection of inspiring bits.
So my question is: how would he feel about a d100 traits table with "meaningful and interesting" entries? I mean, in terms of the density he craves, you can't get much more dense than a tabular format.
I don't think he is being inconsistent at all. An ordinary random table gives you ideas you could have come up with yourself if you had taken the time to do it. A superior table, like a lot of Dungeon Dozen entries, gives you ideas you likely never would have thought of. A designer may have started with the same boring NPC table you use, but has added value by taking the time to interpret that information into an interesting character that you might not have invested yourself. Put 100 of those in a table so you can pull them out on the spot when necessary, and you have a table with high design density (as he is using the phrase).
And if instead of a table, you put a handful of those NPCs in a village, with existing relationships between them, and things they might want from PCs and things PCs might want from them, and you have added even more value by doing that work, as long as it is better than what the DM would have pulled out of his ass by making up a village on the spot.
I get it. I don't buy a lot of modules any more, because even where they are highly recommended I find the new ones are a lot of work to integrate into my campaign. But an orc village or bandit camp, or several human villages that I could drop in whenever I needed to, with interesting NPCs and things already going on in them? I would drop money on that stuff in a heartbeat.
EDIT: also look at his response to one of the comments