Belief in and memory for an event can be independent constructs.

From Clark et al.:

Recent studies have shown that many people spontaneously report vivid memories of events that they do not believe to have occurred. In the present experiment we tested for the first time whether, after powerful false memories have been created, debriefing might leave behind nonbelieved memories for the fake events. In a first session participants imitated simple actions, and in Session two they saw doctored video-recordings containing clips that falsely suggested they had performed additional (fake) actions. As in earlier studies, this procedure created powerful false memories. Finally, participants were debriefed and told that specific actions in the video were not truly performed. Beliefs and memories for all critical actions were tested before and after the debriefing. Results showed that debriefing undermined participants' beliefs in fake actions, but left behind residual memory-like content. These results indicate that debriefing can leave behind vivid false memories which are no longer believed, and thus we demonstrate for the first time that the memory of an event can be experimentally dissociated from the belief in the event's occurrence. These results also confirm that belief in and memory for an event can be independently-occurring constructs.