Inadequate information interoperability in facility management (FM) activities costs time and money being wasted for searching for the needed information from many different data sources. An integrated Building Information Model (BIM), depicting as-is conditions, has a high potential to minimize such wastes. However, facility managers still face the challenge of generating as-is building information models for existing facilities. A main problem with current approaches for generating as-is BIMs is that they mainly focus on capturing and providing geometric information. Many other types of information are missing, such as equipment warranty and technical parameters. The research described in this paper targets the generation of accurate and semantically rich as-is BIMs by leveraging heterogeneous existing information sources, such as drawings and operation and maintenance manuals. This approach was investigated through detailed case studies done in two old academic buildings. Existing information obtained from documents has been compared to industry foundation classes (IFC), COBie, and the data generated from a BIM authoring system. The comparative analysis results reveal several information gaps among different sources.