I don't understand your question. In particular, I'm not sure what you're envisaging can't be done.
I'll try again to explain how I see it, and perhaps that will answer your question.
Scenario 1
The established fiction - let's say - is that the PCs are hoping to find a map; that the map is lost/missing/somewhere unknown; and that one particular PC has just found him-/herself in a study.
Then - let's suppose - the player of that PC, using his/her power to make moves in the game, declares "I search the study for that map we're looking for." The shared fiction now includes the PC searching the study.
The question is: can, and if so does, the shared fiction change to include the PC finding the map in the study?
Scenario 2
The established fiction - let's say - is that the PCs are engaged in a raid of some orcs' stronghold (like, say, B2); that the orcs are very ready to fight back; and that one particular PC, armed with a sword, has just found him-/herself in a room with an orc.
Then - let's suppose - the player of that PC, using his/he power to make moves in the game, declares "I draw my sword and attack the orc!" The shared fiction now includes the PC drawing his/her sword and engaging the orc in combat.
The question is: can, and if so does, the shared fiction change to include the PC killing the orc?
My contention
The shared fiction changes by acts of authorship. In the RPGing scenarios I've described, there are two candidates to do that authoring: the player and the GM.
It is very widely accepted among RPGers that, in scenario 2, the dice are used to mediate the authorship. Rougly speaking, if the dice come up in the player's favour, the player's desire as to the content of the shared fiction is realised: it includes the PC killing the orc. If the dice come up otherwise, then the player's desire is not realised: the shared fiction will include a still-livng orc, and possibly other consequences as well (such as a wounded or dead PC).
There is no technical or metaphysical reason why scenario 1 has to be any different in the way the action declaration is resolved and hence the new ficiton is authored. That is, the dice can be used to mediate the authorship: if they come up in the player's favour, the content of the shared fiction includes the PC finding the map in the study; if not, then the shared fiction does not include discovery of a map, and possibly includes other consequences as well.
If scenario 1 nevertheless is handled differently - eg the state of the shared fiction following the action declaration is determined by the GM reading from his/her notes - then the player does not have agency over that aspect of the shared fiction. Agency in respect of that aspect of the shared fiction has been reserved to the GM alone.
Trying to explain that reservation of agency in metaphysical terms - the map is in <the kitchen, the cave, wherever> and not in the study, and so of course the PC can't find it in the study - is just reiterating that agency has been reserved to the GM. Because the map being in the kitchen (or wherever) is itself just an authored piece of fiction. It's not an independent, objective reality in the way that the actual location of some actual map in the real world would be.
Scenario 2 suffers the same fate as scenario 1. If the GM tells the player he sees an orc in the room, GM fiat can dictate the attack's success. Perhaps the orc is an illusion and the swing cannot connect. Perhaps an invisible wall of force separates the two and the weapon will bounce off with no effect. Perhaps the orc goes first and flees. Perhaps the orc was previously wounded and can offer no reasonable defence and drops immediately.
The issue is the players want to do something using partial information. The DM is privy to more information (perhaps from pre-authoring [its it the bedroom! or the BBEG carries it on its person!], perhaps as a consequence of previous action [the PC's plan were uncovered and the map has been destroyed!], perhaps as a consequence of previous die rolls [the partial failure entering the home resulted in a rival thief fleeing with the map] , perhaps the GM does not believe the party has "earned" this dramatic moment whatever that means, or maybe the party has "earned" the dramatic moment and it's time to move on -- why and how the information has been created is immaterial). The party with more information determines what happens in a manner consistent with what is known as a whole. Dice only come into the adjudication if an appeal to randomness is warranted.
Now is pre-authoring different than improvisational creation in these cases? Probably not from the player's perspective. They know they got into the home and haven't found the map they expected. Is the scene framing going to be different? Maybe. It depends on what the table prefers. Will the play after this scene vary? Probably. But, there are situations where the pre-authoring later scenes and the improvisational later scenes are the same. "The dice said the map isn't here, I know! I'll give it to the rival thief they encountered 2 sessions ago!" vs. "The party is going to kick itself when they figure out the map was stolen by that rival thief they ran into 2 days ago! Let's see what they do when they find the map isn't in the home."
DM-facing games offer players less agency than say FATE offers since they cannot inject declared additions to the world outside their player characters. However, the players' PCs have similar levels of agency. The DM is not restricting the PC agency any more than a fighter's agency is restricted when sees the opponent is 100 feet away on the other side of a chasm. PCs must operate within the constraints of the scene framing and rule set. With pre-authoring, some of the constraints may be initially hidden from the players and are included to provide something for the player to react against when discovered. Other pre-authored constraints are not hidden and exist to provide thematic and genre landscape that the player may choose to -- or not -- engage.
Now, could scenario 1 be determined by dice? Sure! If the parties are operating from the same partial information and the study is a plausible location for the map roll the dice and see if its found. But it needn't be determined by dice any more than any other situation needs to be determined by dice. GM fiat can rely on pre-authoring, gut feelings, sudden inspiration, or listening to the players and agreeing with their ideas. In the end why the fiat happens is immaterial to the players.