Since 2021, the European Union Agency for Railways (ERA) maintains the knowledge-graph-based “Register of Infrastructure (RINF) System”. The RINF knowledge graph is built according to the ERA ontology, mainly constructed using RML mappings on XML data sources, and validated using SHACL. Several applications consume this knowledge graph: users can search for railway infrastructure elements, explore them on a map, and check the compatibility of railway vehicles on a certain route (in this case, using an additional knowledge graph that contains data from the European Register of Authorized Types of Vehicles - ERATV -). Data providers such as Infrastructure Managers or National Registration Entities can upload XML or RDF data, which are consolidated into the RINF knowledge graph, and Railway Undertakings can subscribe and receive notifications when changes are made to some infrastructure elements. Since the initial prototype from 2021, new functionalities have been added and challenges have arisen with respect to the propagation of changes in semantic artifacts, URI design principles for RINF resources, harmonisation of RINF and ERATV reference data, implementation and execution of mappings, and issues in the performance of SPARQL queries and SHACL validations. In this paper we describe the current status of the knowledge graph and applications, and reflect on the main challenges and lessons learned from this development and deployment.