Urban positioning with global navigation satellite systems is challenging because reflections, diffractions, and obstructions introduce multipath interference and non-line-of-sight reception, degrading accuracy to tens of meters. The growing availability of three-dimensional city models offers an opportunity to model these effects. This contribution proposes an enhanced 3-D mapping aided ray-tracing algorithm, RT-based grid weight smoothing and clustering (RT-GWSC). It incorporates a reflection-diffraction model and a positioning scheme using multiepoch weight smoothing and density-based spatial clustering of applications with noise. Static and kinematic experiments in different urban environments demonstrate that the reflection-diffraction model improves sensitivity of non-line-of-sight prediction from 53.7% to 90.7% and overall accuracy from 86.3% to 96.8%. RT-GWSC reduces positioning outliers and achieves accuracies better than 15 m in all tested environments, outperforming raw GNSS outputs by more than 73%, while the conventional ray-tracing approach is ineffective in certain complex environments.