The point spread function of the detector is pretty well characterized, and from there the image is basically just a Fourier transform. It's not as magical as you might think.
Then it looks like they average a bunch of these images to get a maximum likelihood image. What is your issue with that?
Besides, to calibrate on a known celestial body, they'd have to have another micro-arcsecond resolution radio telescope to use. Do you have one they can borrow?