In distributed fusion architecture, processed information in the form of tracked object data is available instead of raw sensor data. Ignoring the cross-correlation in distributed systems by employing Kalman filter in general leads to inconsistent results. Covariance intersection, on the other hand provide conservative results by overestimating the intersection of individual covariances. In this paper, we present a track level fusion by analytically computing the mean and covariance of fused data under unknown correlation. Unlike the covariance intersection method that searches for a minimum overestimate iteratively, the proposed method finds the maximum covariance under unknown correlation. Furthermore, it is proved that the proposed method provides an exact and consistent result in terms of Bar-Shalom Campo (BC) formula. To show the effectiveness of the proposed method, simulation results of Track-to-Track fusion are provided.