The rollout of sixth-generation (6G) networks will support a plethora of innovative services, such as extended reality (XR), which facilitate seamless and highly interactive experiences. These applications demand high reliability and data rates as well as low latency, which presents a significant challenge for the network operators. Recently, reconfigurable intelligent surfaces (RISs) have been proposed as a potential solution to satisfy the stringent requirements of XR communications, primarily focusing on optimizing reflective elements while overlooking system performance in practical urban settings. This work aims to address this gap and evaluate the performance of RIS-enhanced XR services in practical setups. We compare a realistic RIS deployment, which follows the topological characteristics of the considered scenarios, with a deployment that uses the Manhattan grid structure. The simulation results reveal that, as the number of users increases, performance degradation occurs owing to edge server saturation, degraded communication conditions, and resource partitioning. Moreover, Manhattan grid deployment creates uniform RIS positions across different layouts, thereby resulting in significant deviations from realistic scenarios. Our study suggests that the practicality of employing the Manhattan grid to represent real-world scenarios may be questionable.