Uncertainty, whether inherent to the dynamical system under study or due to external factors, is intrinsic to experimental wave tank tests of wave energy converter (WEC) devices. In order to obtain a representative WEC model from wave tank experiments, it is crucial to identify and characterise nonlinearities, uncertainty, and external disturbance affecting the experimental data, to understand the level of data contamination at different operating conditions of the system, and to apply the proper methodologies to compensate for those effects, where appropriate. To this aim, the current study analyses wave tank data from an experimental campaign on a 1/20 scale Wavestar prototype device to identify a range of linear WEC models, using force-input experiments at various levels, with a focus on the identification and quantification of parameters that may cause uncertainty in the WEC model. To this end, the analysed phenomena are classified into electromechanical, hydrodynamic, and structural (support) parts of the test setup. Effective data pre-processing and post-processing steps, to mitigate noise and disturbances, as well as preserving the physical system behaviour, are applied to identifying representative linear WEC models, which serves as staring point for robust WEC controller synthesis. Thus, careful estimation of representative models and a associated uncertainty region overcomes the limitations of a single linear model by covering the complete operational space of the device.