Roentgen Stereophotogrammetric Analysis (RSA) is the most accurate technique to measure micromotion between rigid bodies and it has been widely used in the orthopedic field to investigate lower limb prostheses. The aim of this paper was to investigate the factors affecting the accuracy of measurements performed on the knee joint by the biplanar RSA system. Errors were assessed by means of a test object. A particular goal was to verify the reliability of the system when the patient cannot stand in the ideal position during the examination, i.e. in the centre of the calibration space. Results showed that detection of micromotion is not correlated to the position of markers inside the calibration space. The influence of different materials able to simulate human tissues, both hard and soft, on the data acquisition phase was also investigated. This additional exam was performed to assess whether the results obtained were similar also under as close as possible conditions to the clinical trials. Different experimental conditions were shown to not affect the acquisition phase of RSA procedure; markers were acquired with equal accuracy independently of tissues examined. Since the acquisition procedure is often performed by different operators, the difference between acquisitions made by different operators was compared. This study confirms the reliability of RSA and its accuracy, also in non-standard clinical setup and acquisition procedures.