Accurate and reliable registration is one of the most important issues in computeraided surgery, as small errors may have a large influence on the overall accuracy of the system. The restricted surface-matching algorithm (RSM), initially developed for periacetabular osteotomy surgery (PAO), has been improved to become numerically more stable and reliable. To assess the accuracy and sensitivity of registration, a framework is presented that evaluates two aspects of registration: the sensitivity and raw performance of the registration algorithm are tested in a stand-alone environment, and the integration into a CAS system is analyzed by evaluating the accuracy of the complete system. For the latter tests, spherical-headed titanium screws used as fiducial landmarks provide a reference transformation for the registration. This framework was used to analyze the performance of RSM for PAO surgery. The sensitivity analysis showed the algorithm to be insensitive to noise up to a magnitude of 3 mm. Both the sensitivity analysis and simulated surgical environment tests showed that an accuracy can be attained of better than 2 mm in the region of interest, and better than 4 mm far away from the region of interest. This is sufficient for safely assisting PAO surgeries.