the information is their product.
It isn't. The "reviews" and "published results" are just a happy byproduct of the testing they do for their real business, their software business, and DxO Optics Pro is some kick ass software. They also create embedded software, silicon architectures and optical designs for still and video image real time processing, as well as image quality evaluation, measurement tools, and methodologies. To equate the fact that their web staff may accidentally confuse one english word for another on a site that is merely a free public service (that originates in french, no less), or that the tech writer who occasionally quotes the wrong number is somehow indicative of the quality of their science and engineering staff is laughable. If you were translating your view point into french, would you recognize the difference between "conneries" and a "connard" at a glance? It's like knowing the difference between your argument and yourself.
IF YOU'RE IN BUSINESS AS A TESTING / INFORMATION COMPANY- YOU SHOULD BE DOUBLE / TRIPLE CHECKING YOUR WORK!!
DxoMark.com is separate from the main DxO.com site. DxOMark isn't in the business of testing, that would be DxOLabs. If you think the lab techs are the ones writing the reviews... and plugging the information into the DxOMark site, you are wonderfully naive. Do you think the engineers at GM or Ford are wasting their time plugging in performance numbers on their companies respective websites, or that they double check the websites to make sure the published numbers matched what they tested on the bench, or that they personally make sure that the foreign GM/Ford sites are properly translated? Of course not. DxOLab's technicians have better things to do, and real work to accomplish, like making sure the data they collected translates into real world improvements in their software.
At the end of the day DxOMark is nothing more than a public forum and service that DxOLabs is under no real obligation to provide information to. The fact that they translate their pages as well as they do is amazing, and better than 99% of foreign sites I've visited. The fact is... accidentally getting a number wrong in a product review is pretty meaningless, they aren't selling the products they review, if they were... that would be a completely different matter.
Ever read an auto magazine? They are constantly doing tests, collecting data, etc. I often find that the written article will have at least one discrepancy from the data table. Does that make the people who review the cars idiots? Does that mean that we shouldn't trust them when they say that the 2012 Camaro out performs the 2012 Mustang, that it rides nicer and is more forgiving in corners? Of course not, and although the person writing the (car) article was probably present during the testing... he probably wasn't the one setting up the testing equipment or verifying the numbers.
Jobs get delegated, if you're in the legal field, or advertising, sure, the wrong word or number might get you fired... as there is often a great deal of money riding on it. The same can't be said for DxOMark, which is just a free service that was born out of DxOLabs legitimate business, they have no need to strive for excellence in their web translations, their current incarnation is just fine, occasional errors and all.
they also got the specs wrong on the same line - the 5d3 is actually "expendable"(idiots) down to 50, they wrote "100" in their comparison.
The 5D3's ISO sensitivity for 50 and 100 is exactly the same (tested at an actual ISO 80). The raw files get flagged for Canon's software to process differently, but the way DxO tests the raw files there is probably no difference between the two. Canon software pulls it down a partial stop. Basically Canon's ISO 50 is a software cheat, so who gives a damn. In reality the 5D3 really does only go down to it's 100 setting, which is effectively ISO 80. Long story short, that quote of the 5D3 going down to ISO 100 could have been a mistake, or it could have been deliberate... since the raw files are identical except for some flagging that tells Canon's software how to process the "ISO 50" file.