Software architecture reconstruction helps architects to redocument existing systems and to check code conformance. Existing methods and tools for software architecture reconstruction do not support the structured evaluation of architectural change scenarios based on the reverse-engineered models. We have applied the novel architecture reconstruction method and tools of the EU-project Q-ImPrESS on a large-scale ABB software system from the process automation domain. We have reconstructed a high-level component and connector model as well as component internal control flow from C++ source code. The resulting models can be altered to reflect change scenarios and be semi-automatically evaluated for performance, reliability, and cost properties. Therefore model-driven predictions for these quality attributes in certain change scenarios shall be possible so that architectural tradeoffs can be analyzed. Our presentation focuses on the architecture model reconstruction activities of Q-ImPrESS and describes our experiences when applying the tools on an existing large-scale system.