Which software package are you using? as this may make a difference in specific applications... I know that VENSIM and PowerSim Studio both have build-in methods for autocalibration (AC) of parameter values compared to manually estimation (HC). See: Oliva, R. 2003. Model calibration as a testing strategy for system
dynamics models. European Journal of Operational Research 151 (2003) 552–568. Dr. Oliva provides an overview of how to autocalibate of parameter values, particularly where no or little data exist. It sounds like you do have data but that fluctuations create uncertainty in determining a 'good' calibrated value for your base case. These may be 'structural breaks' in values over time as the system changes (see: http://www.systemdynamics.org/conferences/2014/proceed/papers/P1236.pdf)- this article may provide some method to get you started, and I would advise going through their literature cited, it appears they have provided a good list of references dealing with calibration, validation, and preparation for model testing. I do hope this helps you. Sincerely, Benjamin
Is this fluctuation important to your research? if so, you might consider to build a custom made dimensionless multiplier in a graph to mimic the reality.