In this paper, we propose an improved attribute compression method for point cloud data, using variable bit depth and an error correction table. The evolution of sensors makes it possible to represent 3D models ever more precisely by using ever larger amounts of point data, which makes the need for an efficient compression algorithm ever higher. The emerging MPEG Point Cloud Compression (PCC) standard is trying to meet the demand of compression. We tackled two issues in MPEG-I Part 9 G-PCC. First, because encoding the reflectance attribute with a fixed bit depth limits the compression performance, we propose encoding the reflectance attribute with variable bit depth. Second, because errors can occur between the original and reconstructed values when encoding and decoding the reflectance attribute, we propose minimizing them using an error correction table. Experimental results show that the proposed method outperformed MPEG PCC substantially.