Developing a method to formulate a damping matrix that represents the actual spatial distribution and mechanism of damping of the dynamic system has been an elusive goal. The dynamic stiffness matrix (DSM)-based damping identification method proposed by Lee and Kim is attractive and promising because it identifies the damping matrix from the measured DSM without relying on any unfounded assumptions. However, in ensuing works it was found that damping matrices identified from the method had unexpected forms and showed traces of large variance errors. The causes and possible remedies of the problem are sought for in this work. The variance and leakage errors are identified as the major sources of the problem, which are then related to system parameters through numerical and experimental simulations. An improved experimental procedure is developed to reduce the effect of these errors in order to make the DSM-based damping identification method a practical option.