Because twitter is a shit fest for making nuanced points.

Yes I understand that. which is the absolute 100% point of my original comment that started this chain. Rather than measuring for the card that comes within the mathematical bounds of what they recommend. They should be restricting their bevy of tested cards to those that are current market /especially/ if it means the current market out performs their "ideal" for the price point.
/Yes I think they should be doing market analysis for release date when they do these tests/.

the problem with the way that they do things as "standard practice" is that
1. often the "measuring stick" is out of date by the time the product goes to market.
2. which confuses the consumer because the recommended is the gauge by which many consumers upgrade their systems

so yes while their internal benchmarks are "x setting" at "y framerate" over "z consistency" (for example) sticking your measuring bounds around those so tightly as to not allow the maximum bounds to exceed that goal and not considering market value.
Confuses the ever living shit out of anyone that is a civilian reading the back of a box.

In this /specific/ case. recommending a 780 as a performance gauge /to the consumer/
is recommending
1. a much older card which is already degraded in many modern system situations by the march of software. (e.g the dx9 cap)
2. is insanely expensive due to being out of generation production
3. is out performed by current gen cards which are significantly cheaper worldwide (due to 2.)

TLDR the old standard of "recommended specs" is bad for John Doe because it makes no sense to John Doe. Unless John Doe is already a 700 series user.