There is a basic numeracy issue here on your part lets suppose they only got it 60% correct as per some shelter volunteers/staff... that still leaves pitbulls/bull terrier types as a huge outlier.
That would then be 7 out of 18 though, rather than 12 out of 18, so you're talking closer to one-half rather than 2/3rds.
The point is that there's no confirmation of validation in all cases, even though the number of cases is tiny.
If you can validate each separate case, that's another matter and worth considering, but not until the full data is available.
Then in addition to that this isn't some shelter staff member just saying what they think the dog is... when someone is actually killed by a dog it's a bit more serious/thorough than that and police want to establish whether it was a banned breed etc.
Seriousness is not a factor in determining breed, nor in human ability to ascertain breed based only on looks.
Shelter staff were used in the studies because they have more experience than the average person in identifying breeds and have convenient access to a wide variety of breeds, and because their breed assertions are what people go by when deciding if a dog is safe to adopt.
Whether or not someone was killed or only hurt by the dog is not a factor - If the dog caused harm to any human, it is a criminal offense regardless of severity, and in all such cases the banned status should be established, if only because it adds to the prosecution's case.
You're basically just taking this scatter gun approach in order to throw in some uncertainty then base your entire argument on it, that's led to the really silly argument about the victim not being able to identify a raging ball of fur that fails to note that the victim is dead and now a hasty google to find stuff about shelter staff is a different context entirely!
I've always argued this uncertainty. This is just further substantiation of the reasoning behind it. You like substantiation, don't you? Citations, references?
I'd be interested to hear why you think a victim
or any other eyewitness, in the midst of a stressfull attack, would do a
better job of identifying a breed than an industry professional in a relaxed setting with an ID chart for comparison...
The weight, and reliability, of evidence for breed/ type identification is likely to be higher in fatal attacks than in non-fatal attacks.
Why?
The people who fill out the forms for fatal attacks are the same as those who fill out the ones for non-fatal attacks. The reliability is the same.
Meanwhile you only have some of the fatality reports confirming breed test results. If it's a factor in some, it should be a factor in all, which is why those that don't mention test confirmation are questionable. You'd think that, out of thousands of dog bite incidents, these few
really serious cases would have chapter and verse on every detail.