Some possible reasons:
On the supply side: If narratives are forever unchanging, journalists don't have to engage in difficult (i.e., costly) rethinking in light of new events. Furthermore, they expect that they can afford to dodge a general rethink of a subject because they expect that the the latest event involving the subject won't be in the news for very long. (In other words, the agents in the model understand the theory.) Only rarely does this expectation turn out to be incorrect, but even when it does (say, after three continuous years of bloody war), journalists can still expect to be largely in control of the narrative and thus expect ample opportunity to deflect criticism.
On the demand side: Newsreading is a casual endeavor, not a careful intellectual enterprise. If readers have a casual understanding of an event, group, or person, they may find it difficult or not fun to follow any subsequent articles in which that casual understanding is missing or is demolished. Furthermore, no one likes to be told they were wrong, especially by the people who taught them. This is exacerbated if readers have accepted the narrative and have gone around talking about it to family and friends; they would then have a vested social interest in its correctness.
Any thoughts or other ideas?