This paper discusses the data post-processing aspects of measurement-based load model estimation. Events measured by DFR (Digital Fault Recorder) are used for estimating exponential load models by least squares algorithm. Values are used for comparing the impact of different event filtering approaches. Events are filtered based on voltage and current unbalance ratio, residual voltage of disturbance, voltage deviation, and estimated load model parameter values. In addition, two different methods for calculating representative model are compared.