That seems to be a partern with TWIMTBP Games, AMD run bad on them, Nvidia even worse.....
I think there's definitely a pattern with the last few big releases in the last year or so, but from both camps. Since the new consoles came out, the ported titles to PC have been generally terrible. COD Ghosts (nvidia) and tomb raider (AMD) had extremely poor performance on competing products at launch, which only got better with patches and driver releases. AMD's big release for this year (alien isolation) comes out in 5 days whilst Nvidia have had a few TWIMTBP titles this year, with watchdogs being the worst performer.
Sadly its easier to just blame AMD and Nvidia, then the companies they are sponsoring to endorse their products. Neither company makes the games and all they do is provide assistance to make sure their products work across the range. Optimisation often is forgone to a later date it seems, judging from a fair number of releases now.
With middle earth, shadows of mordor, the game doesn't have an SLI profile yet, but everything else appears to be fine for the settings your GPU is capable of. I run the game absolutely fine on a single 780 ti at the moment with everything on ultra apart from textures, due to my 3gb VRAM limiting me. I don't have an issue with that as it takes nothing away from the gameplay I'm experiencing.
Using results from a deeply flawed benchmark as some sort of indicator of true performance just isn't realistic. The results are skewered massively by the extremely high end max fps caused by the black screen at the start being longer than necessary. The min fps is also no where near accurate, as it depends on how the GPU you own handles its VRAM. The lows that people experience will never even be seen in the game, as it'll be an initial judder when your character phases into the world as the frames are being created and texture information is being dumped into the VRAM.
To properly gauge performance in this game, there should be a time demon in the second zone, which is far more graphical than the first, which only starts 2 seconds into the scene already being rendered, to ignore the super high and super low FPS we see on these benchmarks.