I completely agree, although in my personal experience with Brindlewood Bay it does still feel like solving a puzzle ...
As an aside - I first heard of Brindlewood Bay last night. Due to player illness, my Witchlight campaign failed to run, and so we sat around just talking for a while, and it came up. Really weird coincidence.
That aside - there's some fundamental differences between acts of creation and acts of analysis. Creation can be framed with things the creation must be consistent with (like, say, in writing a sonnet one must be consistent with the rhyme and meter scheme), but the mental processes underway are not the same as analysis. Folks for whom the difference matters will bounce off a creation process masquerading as an analysis process.
But anyway, how about we go into that more granular look at Gumshoe, and how it handles mystery?
Old school games typically place the major burden of a mystery on the discovery of clues. This is typified in Gygaxian searches, in which the player is supposed to specify that they are searching, exactly what they are searching, and how, often in excruciating detail. If you don't pick the right thing to search, or if you don't get a good roll, you don't find the clue.
In 3e, they lightened this up a bit. The search action declaration is more broad - "I search the room, square by square." No need to specifically note that you are searching the leg of the bed, that's assumed when you get to that square. But still, if you fail the roll, you don't find the clue.
This necessitates things like the "Three Clue Rule" - the GM has to assume that the players will get some bad rolls, and will miss clues. So, if discovering the clues is to be a real option, you need more clues. This is less about railroading than about plain statistics, but I digress...
Given that this can be a little tedious, and given that most GMs are not, in fact, seasoned mystery writers, there's a tendency for the clues to be pretty blatant - the clue is a scroll tube of letters between the Duke and some unknown correspondent, detailing when they are planning to kill the king, or the like. When you go to the appointed place at the appointed time, you encounter and fight the unknown correspondent, defeat them, and unmask the... Duke's wife! Or whoever. Mystery solved!
And, yeah, the results are kind of linear, and are contingent on a specific resolution, so they can feel railroady. Granted.
Gumshoe looks at satisfying mystery fiction, and notes - rare indeed do the investigators
fail to find the clue. Spock doesn't scan for life forms, fail, and go, "Gee, Captian, I'm sorry, but for some random reason I can't scan my butt with both hands today." If there are life forms, Spock finds the darned life forms! What the scan doesn't tell Spock is
why the life forms are relevant. Gumshoe proceeds based on the idea that
searching for clues is actually kind of tedious, but
thinking through their context and meaning can be very interesting.
So, in Gumshoe games, finding clues is easy. If the character has the right skill, in the right place, and asks, basic relevant information is given, no roll required, no chance of failure. If there's more information, they can spend a skill point to buy it - again, no chance of failure, only a question of how much they want to invest on getting one set of information.
Without having any concern about what information the players will be able to find, the GM is free to distribute it more widely, in smaller bits that provoke more thought.
With the focus now on thought, rather than search, we are more solidly in the space of exploring how the characters relate to the issues - playing to find out what this means for them, and how they want to proceed with what they discover.