The Terahertz (THz) band (0.1-10 THz), which supports Terabit-per-second (Tbps) data rates, has been envisioned as one of the promising spectrum bands for sixth-generation (6G) and beyond communications. In this paper, an angular-resolvable wideband channel measurement campaign in an indoor L-shaped hallway at 306-321 GHz is presented, by using a frequency-domain vector network analyzer (VNA)-based channel sounder. Four line-of-sight (LoS), six quasi-line-of-sight (QLoS) and eight non-line-of-sight (NLoS) receiver points are measured. However, measured data spreads due to the rich scattering environment and the antenna pattern, which puzzles traditional clustering algorithms. To solve this problem, a simulation-assisted Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clustering algorithm is proposed, where the deterministic simulation result is extracted to adapt the conventional DBSCAN algorithm. The proposed algorithm outperforms conventional clustering algorithms like DBSCAN, K-means, and K-power-means in terms of Silhouette, Calinski-Harabasz and Davies-Bouldin indices. Furthermore, the THz multi-path propagation in the L-shaped hallway is elaborated, and channel characteristics of multipath and clusters are analyzed in depth.