High-speed PCB design for signal integrity (SI) is about feasible material selection, trace geometry determination and optimization of discontinuities, where the accurate PCB material characterization is essential since incorrect material properties may lead to misleading results and wrong design descisions. The previous studies have revealed that the simulated time-domain reflectometry (TDR) impedance in material characterization, which is based upon the transmission-line-based methods, is erroneous when compared to the measured value, although a good agreement between simulation and measurement in the frequency domain can always be reached. In addition, it is also shown that achieving a satisfactory correlation in both transmission phase and trace impedance is a challenge for SI engineers. This implies that the transmission-line-based approaches, which are widely used in industries, are not perfect and that the extracted PCB material properties are not accurate enough. In this paper, a step-by-step investigation is performed and demonstrated to disclose the root causes of the TDR impedance discrepancy. It is found that the disagreement in TDR impedance is contributed by multiple factors which need to be taken into consideration during material characterization. The improved simulation result exhibits excellent consistency with the measured trace impedance. The suggestions to hardware designers on how more accurate PCB dielectric properties can be obtained are given by addressing the TDR impedance discrepancy issue.