A laboratory well-mixed, constant supersaturation, semi-batch gibbsite precipitator, with clear liquor advance, was modelled using a discretized population balance (DPB). This experimental configuration is inherently non-stationary (i.e. does not reach steady state). The model predictions of the transient crystal size distribution (CSD) were found to give good agreement to the experimental CSDs initially but deviated apart with time. The cause of this deviation was identified as error propagated in the computation of the DPB model. The source of the propagated error was investigated. The error contributions from the growth and agglomeration term discretizations were found to be small compared to the magnitude of the observed deviation. Monte Carlo simulation demonstrated that the main contribution to the observed error is from uncertainty in the estimates of the agglomeration kernel and growth rate parameters, which are estimated from experimental data and used in the simulation. The findings highlight the need for more precise kinetics estimation procedures and advocate care when simulating the CSD of non-stationary precipitators over longer time scales, which may have implications for precipitator start-up simulations.