The rapidly growing DER penetrations require more accurate distribution models including more granular load models. While AMI and other modern distribution measurement sources provide a new level of visibility to distribution load modeling, there is limited consensus in how these data sources should be efficiently utilized. This paper examines conventional load modeling concepts and methods on a large-scale real utility AMI data set. The commonly used load factor and diversity factor concepts are analyzed with respect to various sensitivities such as customer group size. Additionally, errors of the commonly used transformer kVA and kWh load allocation methods are investigated. On the analyzed data set, transformer kWh allocation clearly outperforms kVA allocation independent of the allocated feeder load time or level. This research demonstrates the additional insights that be achieved by integrating AMI and other modern distribution measurement sources with the distribution planning tools, such as OpenDSS or CYME.