Rather than masking outliers in the residuals image (data minus model)---right before performing ellipse-fitting of the central galaxy---mask based on the custom Tractor catalog, which is more conservative.
Or maybe we can just use the model image. Here's one example (excluding the central galaxy). We should be able to get a robust sigma for the image and then just threshold mask (+dilate).
Rather than masking outliers in the residuals image (data minus model)---right before performing ellipse-fitting of the central galaxy---mask based on the custom Tractor catalog, which is more conservative.