Hmm ... it has bothered me that running 4% under the minimum gives a 5000% increase in measured flow.
(Yes, I would have selected a model with the expected maximum flow comfortably below the maximum of the meter.)
The "Typical error" diagram shows that it measures flow at MINUS 3% just under the minimum. This probably applies to all models.
The full/empty pipe still has to be resolved. (Was the stain produced during the test, or for example while it was emptied for shipping). I'd love to see an accurate pipe diagram to see if it is topologically possible to run it with a non-full pipe.