Sorry, I'm a bit late, but I think Juraj has already explained everything nicely:
Corona Shadowcatcher is intended for this exact scenario. You have a real-life footage (could be still or sequence) you want to insert some 3D objects on top of that sequence, and make sure that the real-life objects capture shadows, contribute to the GI, are visible in reflections, refractions, etc.
So if you would take your real-life footage, set it as your viewport background, model the real-life building, and assign the shadowcatcher material to it, then that 3D building would now catch shadows, generate GI, etc based on the 3D scene (lights, other objects) and the color information in it. That's a simplification, but basically the colors from the real-life footage would act as diffuse colors for the 3D model.