Distribution networks have experienced significant changes in the last two decades, and have moved from passive networks to active ones. These critical changes require careful control and monitoring, involved with measurement devices and communication platforms. However, in distribution networks with a very large number of customer loads, it is not cost-effective to consider extensive measurement devices. As a result, distribution state estimation (DSE) methods can be used to estimate the states of the unmeasured nodes, where pseudo measured data is used for unmeasured nodes to make DSE method observable. However, pseudo data brings high uncertainty to DSE algorithm, which decreases the accuracy of the estimated states. To increase the accuracy in pseudo data, conditional multivariate complex Gaussian distribution (CMCGD) is employed in this study. CMCGD method increases the accuracy of pseudo data in presence of high correlation between the sets of aggregated load data. As a result, loads aggregation is considered in this study to decrease uncertainty and increase the correlation in pseudo data. Although loads aggregation increases the correlation, it decreases the size of the impedance matrix by reducing the number of branches. Ignoring the cable impedances in distribution networks, will decrease the accuracy of the estimated states, even with a highly correlated set of pseudo data. In this paper, we aim to show that each distribution network with its own cable impedances requires a balance between loads aggregation and the reduction in the size of the impedance matrix. An investigation is provided on a typical distribution network to show how to achieve a minimum estimation error, when both factors of loads aggregation and size of the impedance matrix are considered.