On Rationalizing Expectations Using Rank-One Updates of the Kalman Filter