There's a perfect example. The encounters all pretty much follow one after the next. Until you solve X encounter, you cannot move on to Y encounter. Until you find the secret door in the pit in the first room, you don't get to move on, for example. The few branches you get are almost all dead ends (in more ways than one.)
I was going to bring this up last night but I was pretty tired:
If you're modeling the geographic flow of the Tomb of Horrors, you cannot ignore the teleport effects.
If you just look at the layout of corridors, the Tomb of Horrors is a fairly simple branching dungeon. But with the teleport effects there's actually quite a bit of sophisticated looping going on.
And this analysis assumes that other solutions aren't possible. For example, a couple of friends of mine successfully navigated the tomb by
casting locate object on their personal items after they were teleported to the demi-lich's tomb and then using stone shape spells to tunnel directly there.
That being said, I agree with your general points.
OK. Now we are talking scenarios. In a mystery situation the players begin with something to solve and some introductory information to start the investigation. This could be linear if the mystery has only a single way to solve it by following a prescribed trail of breadcrumbs-but that would also make it a railroad.
I would disagree. Let's assume a simple mystery structure of "solve for the location". In other words, the clues in Location X point to Location Y; the clues in Location Y point to Location Z; and so forth.
Even if we follow the Three Clue Rule and include multiple, diverse clues at Location X pointing to Location Y (allowing the PCs to approach Location X in a number of different ways), I still think it's fair to describe this X to Y to Z structure as being a linear one.