Having been throug at least a half dozen attempts to reword the text prompt, including the use of scientific names, it becomes VERY apparant that the fault in fact IS the images the AI is trying to use are quite simply put "Only REMOTELY "close to" matching"!
Various wordings for the basic idea "American skunks sat at a table eating waffles" has resulted in variations of minks, ferrets and badgers, bur NOTHING "like" skunks, striped skunk (Mephitis mephitis), eastern spotted skunk (Spilogale putorius) or American hog-nosed skunk (Conepatus leuconotus).
So the idea would be the ability to be able to upload whatever the ACTUAL image of the desired object is that the AI is SO DOGGEDLY GETTING WRONG so that it could be added to either some personal image database ot the greater image database