One of the things which has often perturbed me a little about the so-called "Story Games" has been how they shy away from really getting to grips with the mechanics of the genre form they're going to try and emulate.
For example, A Wilderness of Mirrors brings onto the table the concepts of planning a mission and Control eliminating a "rogue" agent - but it never ever talks about what kinds of things happen on a spy mission and why those things happen that way. This means on the one hand that the game is totally flexible - but it also means that unless you can bring substantial genre understanding to the table yourself, things may not go well at all. The game works beautifully in the hands of people who "get" the genre really only needed the gentlest of prompting. The kinds of people who probably don't even really need the system to begin with. Examples are lamentably common, from Zombie Cinema to Fiasco.
On the other end of the spectrum, probably the majority, there are games that focus on the minutae of procedure that are the result of these story/structural forces - they are, in essence, a recipe book rather than a manual of principles. They allow you to replicate an experience, but not necessarily understand why. I want to talk a bit more about two specific games that fall into this category: A Taste For Murder and Dirty Secrets, and I'll try and come back around at the end to talk about how to cover both ends of the problem to provide a story schema without straight-jacketing the gameplay.
Both games try and evade the central question posed in all detective fiction: Whodunnit? Both games explicitly want this to be unresolved until the last possible moment, when it is resolved by the mechanics and then post-rationalized by the play-group. Normally in detective fiction, "Whodunnit" is approached through 4 different and contra-balanced questions: Motive, Means, Opportunity & Alibi. These are the basic story-mechanical entities in play, and the basis of almost all classical detective fiction is that one of these things is never directly stated, but must be rationally deduced.
This approach is the opposite of Columboism - instead of everyone knowing the murderer at the start, nobody does, and indeed, every character is Schroedinger's Murderer: there's no way to know until you crank all the way through to the end of the system and open the box.
This evasion of the central question seems like a neat circumvention of the well-known problems with investigative games, but actually, this evasion means that these two games leave only the surface appearance of the genre intact. In doing so, the game must inherently violate almost all of the famous "Rules of Detection" that lurk in the story mechanics of most of the "Golden Age" fiction that inspired A Taste For Murder must obey. With reference to S. S. van Dyne's specific fomulation of the formula, in effect, the core principles of "ratiocination" (i.e. logical deduction) are thrown out on the simple basis that there is no logical process for determining the guilty party.
What that means is that inside the fiction, we no longer need to really address the four principal questions of detective fiction. The action of the game can thus be nominally "investigative" but the investigation is actually irrelevant. It can be completely impossible to solve the crime in fictional terms: without the arbitrary and irrecovable intervention of the system, ambiguity could always remain. That makes the stories inherently unsatisfying as mysteries - it shunts the games firmly into another interesting genre, the melodrama, where the revelations and counter-revelations are given impetus by a murder.
I think this is reinforced by the play advice after the main rules, which is all about the interpersonal relationships and period detail. For me though, this begs the question of the game's existence. Why would I play a game called A Taste for Murder for my dose of melodrama and interpersonal strife?
Dirty Secrets is even less functional as a game, primarily because where ATFM calls for a collaborative "and" approach to improvised detail, Dirty Secrets actively and enthusiastically calls for blocking and obstruction at every turn. It is the most adversarial game I've ever read. It does redeem itself slightly by mentioning the key phrase "Theory of the Crime." What it says is that in every scene, each player should have a theory of who did it, why and how, and be working to insinuate facts which support their theory into the game. It gives no guidance on what this theory should look like - what are the moving parts, what are the constraints - but it does mention it. Once you as a group decide that the adversarial aspect isn't necessary, and collaborate on the theory of crime, moving the conflicts into the fiction rather than around the table, the game does produce a fictionally-functional crime, inasmuch as you are capable of it based on your understanding of genre.
Is it practical to demand any genre-emulating game to also function as a poetics for that genre? Do we need A Taste For Murder and Dirty Secrets to spell out the possible crime schemas? Well- yes. I think that there is a tendency, exemplified by these two games, to try and treat the game mechanics and the fiction they create separately. The two are usually explained in parrallel, encouraging the reader to see the points of equivalence, but there is a strong causal relationship to story "anomalies" from certain mechanics. What we're looking for here is a statement like "X Rule implies Y player behaviour."
In A Taste for Murder, for example, all the play advice is structured around the melodrama, but there is no specific discussion of how the need to have motives for murder will shape that melodrama. The advice is "investigate relationships, don't investigate clues." But there isn't a discussion on what kind of relationship is appropriate for the needs of the story.
One game that does this fairly well is EPOCH. It is built around explicitly calling out the effect of certain mechanics on player behaviour, and hence on fictional outcomes. For example, when your survival inside the fiction relies on you being interesting, then you make that effort to be interesting. In the play-advice, EPOCH is one of the clearest games around for outlining what will actually happen at the table, and how the game mechanics push the game to turn out that way.
For my money then, we still need a better mouse-trap. For all that there have been innovations around the field, the fundamental problems that exist with investigation-based games have not been resolved. And before anyone says "Gumshoe" - that system does not address the story mechanics of a crime or its investigation, it just takes the random component out of it by guaranteeing clues will be distributed during the game.