This sort of reconstruction problem from VLBI measurements is under-determined so you need to insert priors/regularization to get anything at all. The priors in this case are pretty weak (from a quick read of the CHIRP paper). https://arxiv.org/pdf/1512.01413.pdf