In a recent paper, we proposed group-by-group bit filling (GBF) and group-by-group bit removal (GBR) to boost the efficiency of bit loading. Generally, GBF/GBR converges to a suboptimal bit-allocation profile (BAP), and thus, a supplementary greedy bit-filling/bit-removal procedure is required to converge to the optimal BAP. In this correspondence, we propose refined GBF (r-GBF) and refined GBR (r-GBR) to overcome the imperfection of GBF and GBR. The key idea is to adjust the group size dynamically such that the optimal BAP can be directly achieved for any available target data rate. Based on r-GBF and r-GBR, a couple of bit-loading algorithms, which are named iterative GBF and iterative GBR, are presented. Numerical results show that the newly presented algorithms converge to the optimal BAP properly and outperform existing optimal algorithms in computational cost.