In plan-based interactive narrative systems, user actions must be integrated with an ideal story structure in order to strike a reasonable balance between user agency and story quality. A process called mediation addresses this issue by selecting appropriate responses to possible detrimental user actions in a given interactive narrative plan. One way the algorithm mediates exceptional user activity is accommodation, the process of replanning future events to incorporate an unwanted action and accomplish the story's authorial goals. However, this approach is limited by an underlying assumption that the story world exists as an object outside the user's subjective experience. In this paper, we incorporate a model of user knowledge into plan-based mediation in order to widen the space of plans the algorithm may search during accommodation by allowing the process to replan past events that occur outside a user's imperfect knowledge of the story world.