The article said basically what I tried to explain… if the breeds of the dog aren't in the data base, you are pretty much out of it. Also, btw, if you read the history of the development of the amstaff, terriers are in it... so the make up also makes sense.
I guess if you think the test will well you 1 part GSD, 2 parts chihuahua, 1/2 part fox terrier and 1/2 part Chow, you will think they are awful because they don't. They can give guidance and tell you the dominant breeds if those breeds are in the data base. And the more mixed, the far less likely to get much useful. So I guess I will agree and say it has limited use for any very mixed dog and for those not in the database. But I know several people who KNEW the parentage of their mixed dogs, sent in for fun and got accurate reads. But these were dogs with only 2 breeds mixed, not several breeds.
So if you have a dog that the breeds you suspect are in the program, I'd say worth a try.
http://www.dogbreedinfo.com/articles/dnacaninebreedidentifcation.htm
I just find it fun, and the more people who test, the more samples they have, the better they will get. New technologies take a while to perfect… so understand the limitations and use the company that has your suspected breeds.