Lore wouldn't mean much for a game world if you didn't follow it. If you're using an established game world with expectations from fans, you should follow the established lore as much as possible. An occasional exception such as Drizz't or the like is acceptable, but in general the lore and the actions of the creatures in an established world should match. If you don't like it, write a world of your own and establish the lore you wish. That's how I see it. When I'm playing Forgotten Realms or Greyhawk, I want the world to feel like those worlds. If the DM wasn't following the lore of the world, it wouldn't feel authentic. I'd rather he just change the name of the world and establish the lore he wishes.
Okay, let's spool this out to the end. And rather than a
reductio ad absurdem thought experiment, I'm going to hold up my for-real, currently-in-progress home game as a subject to test your assertion.
I run a campaign in the Forgotten Realms. As I mentioned upthread, I got buy-in before we started play to depart from the tracks of canon, so I established that this was an AU timeline set in the year 1501 where the Spellplague never happened. I decided my approach to lore would be that the 3e FRCS was more or less historical truth unless it would be more interesting if it weren't, and gave myself all of the time between then and our current year to plant setting-disrupting bombs. I broke off the dynasty of the Obarskyrs, because I wanted a Cormyr in the background that had been partially destabilized by an interregnum and I wanted a house of my own invention on the throne. I centered the campaign in Westgate, but established that a generation before, adventurers had cleared out the Night Kings and put the city under the rule of a Lord Steward, because I can take or leave Manshoon (uh, spoilers, I guess) but wanted a new criminal underground that was a front for a Great Old One cult, and because I liked the idea that the city was in tension between factions who welcomed the new order and those who resented having a king-in-all-but-name in charge (including the Croamarkh, now stripped of most of his traditional power, and any of the old noble houses he could make his allies). And I let the effects of that ripple out all across the Dragon Coast and the Sea of Fallen Stars, and dropped my PCs half-unwitting into that powder keg and let them run around breaking stuff.
It still feels like the Realms to me (though I confess to being innocent of a lot of Realmslore before I got started) - it's got the factions, and the gods, and Elminster himself, who has even popped in for a couple of cameos. But it's not just the "occasional exception" that I changed - I tossed out a bunch of things that probably feel like comfortable furniture to people who are super-familiar with the setting. Yeah, the canonical lore was my springboard, but it was only where I started, and I twiddled a
bunch of knobs on that sucker to make it serve my needs. I am most definitely not "follow[ing] the established lore as much as possible" - I'm following it for as long as it's fun and supports the campaign I have in mind, and then I'm doing what I d!mn well please with it.
So my question is: Are you
really saying that if I was going to do all that, I should've just invented my own world instead, or filed off the serial numbers and changed all the names? Even though that would've exponentially increased my workload? And please note that the question isn't whether or not you'd enjoy playing in my game; it's that, given that its intended audience seems to be having an outrageously good time, are you actually making the prescription for my table that you seem to be, and if not - in light of examples like my own - what is it that you do actually mean to say?