5G Non-Public Networks (NPNs) for industrial networking focus in particular on reliable low-latency communication to enable new use cases. Traditionally, the network itself as well as the end devices are black-box, closed-source platforms. This prevents a fine-grained analysis of the influence of the different components, when evaluating the end-to-end One-Way Delay (OWD). Recently, open-source 5G Radio Access Network (RAN) implementations reached a certain level of maturity allowing a higher separation of the components. In this work, we perform a comprehensive delay analysis with a focus on the User Equipment (UE) induced impact. Our results show that devices containing the widely used Qualcomm X55 chipset introduce an additional delay of up to 6ms, compared to the successive Qualcomm X62. This puts a lot of previous 5G related measurement campaigns into a new perspective, as the X55 was the first chipset allowing the Stand Alone (SA) operation.