We compute a pseudo-dataset by Monte Carlo simulations featuring important characteristics of US agriculture, such that the initial technology parameters are known, and employing widely used datasets for calibration. Then, we show the usefulness of this calibration by applying the duality theory approach to datasets bearing as sources of noise only the aggregation of technologically heterogeneous firms. Estimation recovers initial parameters with reasonable accuracy. These conclusions are expected, but the proposed calibration sets the basis for analysing the performance of duality theory in empirical work when datasets have more observed and unobserved sources of noise, as those faced by practitioners.