This paper analyzes the impact of specular signal reflections on the accuracy of Received Signal Strength (RSS)-based localization for Internet of Things (IoT) devices using the weighted least squares (WLS) regression algorithm within a two-ray propagation channel. Simulations with realistic WiFi/BLE settings, considering distance, antenna heights, and carrier frequency, reveal that localization accuracy is significantly influenced by deep fades caused by surface reflections, which depend on the geometry of anchor-target positions. A pseudo-outlier elimination approach based on feasible localization distances effectively mitigates this issue, significantly reducing localization error. These findings offer practical insights into the performance of WLS-based IoT localization in two-ray environments and lay the groundwork for GPS-free or GPS-denied localization systems in challenging scenarios, such as overwater environments, where two-ray propagation is predominant.