Many old unreinforced masonry (URM) structures still in use need to be assessed considering the safety requirements proposed by current codes. Because of the complexity of the URM response, sophisticated numerical descriptions are required for an accurate structural assessment. When inverse analysis is used for the identification of material properties, the study of the effects of measurement errors is essential for assessing the robustness of the adopted procedure. In this work, inverse analysis techniques utilising Genetic Algorithms are employed to calibrate elastic material parameters of an advanced mesoscale model for URM. In order to apply this strategy to in-situ low-invasive investigations, a non-conventional flat-jack test setup is proposed. The potential and limitations of the method are analysed using computer-generated pseudo-experimental data with different noise limits. This allows the evaluation of the influence of the measurement equipment precision on the stability of the inverse problem.