Now we know that Tamron may have autofocus issues right out of the box, and may require tweaking via the tap-in console...
The-Digital-Picture does their lens tests using careful manual focus, so AF performance does not affect the image quality samples at TDP. How much coffee Bryan has had before testing a lens might!
The real users on the Internet, on the other hand, praise the lens, and certain reviews even include sample pictures that indicate decent sharpness.
Lenses can look pretty "decent" in actual photos and still look pretty bad in magnified crops of different areas of a test chart. Flat test charts are used to test lenses precisely because they show every minor flaw a lens has.
Lenses that demonstrate field curvature will look blurry in the corner of test charts not because the lens is actually that "soft" at the corner but rather because when the lens is optimally focused at the center of the test chart the lens is focused slightly in front of (or behind) the edge of the test chart.
Some real world reviews indicate that buyers return as many as three Tamron lenses before receiving a decent one.
Tamron has always had a reputation (rightly or not) that there is a wide variation from one copy of a lens to another in terms of manufacturing tolerances that affect lens performance. Many of their older and cheaper lenses do not include any internal adjustments to correct for slightly less than optimal alignment of the various lens elements. Whatever the result is of combining all of the slight variations of each part due to manufacturing tolerances, that's pretty much what you get with such a lens. Their newer, higher priced lenses do have adjustment points but they're not always dialed in well from the factory. Canon tends to put at least 1-2 adjustment points in most of their lower cost lenses, and even more in their more expensive lenses.
It may be that TDP tested a less than stellar sample of the Tamron lens. In comments he made elsewhere regarding a review he did years ago for the Tamron SP 17-50mm f/2.8 Di II VC Bryan stated that he thought he had a bad copy of the lens and he hoped to examine another copy and complete the review. He apparently never did, since the review has never been completed. He has made such accommodations on more than one occasion for other lenses when a second copy performed much better than the first. In contrast to the results with the Tamron SP 17-50mm f/2.8 Di II VC, TDP's review of the Tamron 17-50mm f/2.8 Di II (earlier non-stabilized version) showed the copy Bryan reviewed was a very good lens. Most other reviewers and testers said the non-VC and VC Tamron 17-50mm f/2.8 APS-C lens were very similar in optical quality. DxO showed the VC version was slightly sharper than the non-VC version in their tests. I've been very happy with a copy of the SP 17-50mm f/2.8 Di II (non-VC) I bought in 2009.
As Roger Cicala at lensrentals.com said back in 2008, "Now don’t get me wrong, there are bad copies of lenses out there, as best we can tell ranging from 3% to 7% of lenses. And we know, despite our checkout procedures, that 1 of 400 lenses or so will be damaged in shipping and arrive not functioning." That number does not include lenses that customers use and return at the end of their rental period that then fail the lens performance tests lensrentals does for every lens between every rental.
By 2011 Roger has reduced that percentage of "bad" lenses from the factory. "Whether it's quality control at the factory or getting knocked around in shipping our experience is about 2% of new lenses need to be exchanged."
Just because a specific copy of a lens is "soft" due to misalignment does not always mean it left the factory that way.