After you have created a 3D model, you can select "Build Texture" under "Workflow". There, you can select the mapping mode "Orthophoto". After creating the texture, select "Tools" --> "Export Texture".
As far as I can judge, the resulting image is an orthophoto (without georeferencing). I have tried this with legacy aerial photographs (from 1944, n=20) and also with some ground based images (n=33). Using only a few reference points, the orthophotos can be georeferenced, and they align well with Google Earth.
Now, the interesting thing is that in the standard edition I can not set ground control points. The 3D model is in an arbitrary coordinate system, with the X, Y and Z axis definitely not aligned with real world coordinates.
My question is how PhotoScan choses the plane on which it projects the input images to create the orthophoto. Does it simply use the plane defined by the maximum extents of the model? Does it use the bounding box? Any ideas out there? Anyone from Agisoft who wants to tell me?