Thank you for asking first and not just running a potentially bad data set over again. You wouldn't believe how many people just submit the same data over and over and then want a refund when it doesn't turn out... What's that definition of insanity again?
You used 80% overlap and did a full grid mission which means that everything is basically covered 50 times. That is way too much overlap. The southern portion that didn't turn out well was covering an area with lots of motion in it. If you have 50 images of something but they are all different the system has a hard time locking on to features to reconstruct the scene.
We will grant a one-time refund for this. The job will likely turn out fine if you just use half of the data (one set of passes, not the cross passes of the full grid). 80% is fine for that survey area.