Accurate seismic vulnerability assessment of existing residential buildings requires numerical models capable of reproducing their actual dynamic behaviour. In this context, finite element (FE) models often suffer from significant epistemic uncertainty due to simplified assumptions regarding mass distribution, stiffness contribution and boundary conditions. This study presents a methodology for the calibration of FE models of residential buildings based on insitu dynamic measurements, with particular emphasis on the influence of non-structural elements and staircases. The methodology is applied to a 10-storey isolated reinforced concrete residential building, representative of common mid-rise housing typologies. Ambient vibration tests and operational modal analysis are used to identify the fundamental dynamic properties of selected buildings, including natural frequencies and mode shapes. These experimental results are employed to iteratively calibrate numerical models by adjusting material properties, stiffness contributions and mass distribution. Special attention is paid to the role of non-structural components, such as masonry infills and partitions, which can significantly increase global stiffness and affect modal properties, as well as to staircases, whose structural continuity may induce substantial changes in modal behaviour and load transfer mechanisms. The calibrated models provide a more realistic representation of the measured modal behaviour of the buildings, thereby reducing epistemic uncertainty in stiffness and mass modelling. This modal-consistent numerical baseline enables more reliable subsequent seismic analyses (linear and nonlinear) and, consequently, a better-grounded seismic vulnerability assessment of existing residential building stock.