I havent seen anyone try to replicate AMDs results in game v 980ti yet. I had a check of some of the reviews and most (All I think) had AF maxed out. You will notice that AMD choose not to have AF on in most of those game settings. AF seems to have had quite an impact on the performance of the FuryX in the games listed below in the reviewers benchmarks.
So the results AMD listed are not what I would call inaccurate, but it is a very selective approach, one that should have been blindly obvious would backfire.
When AF is maxed the 980ti seems to have a smaller decrease in performance that the Furyx, resulting in a small percentage lead.
Also about the 3dmark score it doesnt say whether the AMD results are graphics only or complete system. I think it is graphics only as other system components can lead to a differing result. 4007, could be the forbes journalists total system result, which is extremely close to Guru3d's result. And if you look at the Guru3d result the image shows they are referring to the total system benchmark
http://www.guru3d.com/articles_pages/amd_radeon_r9_fury_x_review,25.html
After all is said and done I am in two minds about the Fury X. I dont consider it a failed product like many others see to do. But it is quite expensive, and it honestly feels like the prelude to the main event, 14nm GPU's and HBM2. Perhaps the following months if voltage control is unlocked and drivers are improved we might see a bit more potential (A price drop would be nice too)