GNSS/INS-RTK (GIR) systems are usually used to generate high-definition maps by accurately estimating the vehicle trajectory during the data collection phase. This strategy can be applied in open-sky environments and simple road structures. On the other hand, the satellite signal quality becomes very low in challenging environments such as long tunnels, high buildings and dense trees. Consequently, relative position errors are generated in the maps, especially in the revisited areas where there is a high potential to combine data with different global accuracies. This study illustrates the low performance of GIR system to generate accurate maps and mainly analyzes the corresponding influences on a probabilistic LIDAR-intensity-based localization system in a challenging environment in Japan. Furthermore, these effects are verified by comparing the experimental results with a precise map and demonstrate the relevant influences using a unique Graph-SLAM framework.