(Note: If you're coming to this cold—say, from work on Monday morning—you might want to read the post below this one, "Hypothetical Technical Question," before you read this.)
To forestall further comments (there are already a lot), the proper answer to the question in the previous post was given by Edward Taylor and PhysicsMan, and implied by only a few others: in PhysicsMan's words, "The device's total score should be the max of its scores in any sub-field." So, Device 1: 9; Device 2: 7. Device 1 has higher image quality. If you think it doesn't, then try getting a quality of 9 out of the second device. We'll be waiting.
To make this more intuitive, consider a more concrete example—say, two lenses being ranked for resolution. The first lens can resolve 200 line pairs in the very middle of its field, but resolution quickly falls off to 100 halfway to the edges and much worse than that in the corners. Lens 2 can resolve 150 line pairs across its entire field. If you were doing surveillance photography and needed the utmost resolution to be able to, say, read license plates from great distances, which lens would be more useful? A single-metric score doesn't describe either lens very well, but lens 1 still has better resolution, absotively and posilutely.
That, anyway, is (to my understanding) how dxomark.com is doing it, because they've decided to measure the highest potential image quality of each device they look at. If it were up to any individual to make a single number out of a set of data, they could pretty much decide to do it however seemed most reasonable or useful to them. There are many ways to do it, as we saw in the comments.
If the two devices in my hypothetical question were cameras, there might be a valid reason for preferring the device with the lesser image quality—namely, that its average quality across all settings is higher. "Adirondack Pete" had it very right when he said that if I were reviewing both devices, I might well recommend Device 2 after going on and on about all the subtleties!
Hypothetical questions (a subset of what I think the Germans call Gedankenversuch, thought experiment) can be very useful in debating, because they can usefully isolate factors that are commonly confounded with other factors in real-world situations. Useful as they are, however, I actually consciously try to stay away from hypothetical questions in my writing because so few people seem to be able to cope with them. One of my (many) casual interests is reading about the typical fallacies of human thinking (you might begin with this nice little book, or this one)—one lesson of which is that humans in general definitely don't like hypothetical questions. It goes against the way our brains like to work. (We're especially leery when we suspect the hypothetical question has been deliberately framed to bolster one argument and discredit another.)
Typical human-brain* answers:
They're both Republicans and I'm a Democrat, so I say neither.
Hoover, except when he's sitting down and Bush isn't.
The question can't be answered.
If Hoover were 9 feet tall, then he could dunk. I've never read anything that said any president could dunk.
You haven't given me enough information. I need to know their shoe sizes and how big their noses were.
George Bush was better looking.
Hoover > Bush for the stated premises.
You have to rephrase the question so that Bush is 5'6" and Hoover is 5'8" and how much their wives loved them was the sole determinant of greatness.
Depends on how you look at it; the relative power of the presidency in Hoover's day was less, and the situations he faced were very different...[etc.].
Which George Bush?
We should all just try to get along and stop arguing about who was the greatest this or the greatest that.
[Long essay about measurement standards down through the decades, citing shifting tides in the definition of what a "foot" is and naming historical sources that claimed Hoover wore 6-inch platform heels and paid his doctor to exaggerate his height, etc., etc., etc....]
The only possible answer to the above question is that Hoover > Bush for the specific premises stated, and you properly understood the question if you realized that it has nothing to do with Bush or Hoover or great presidents; the point is to be able to accept the imaginary premises and argue logically from them. All the other answers are nonresponsive and and/or beside the point. (In my hypothetical question below, any answer that took the form of "1:X, 2:Y" where X and Y are integers between one and ten would earn at least half credit...or perhaps full credit, depending on what the examiner was looking for...for instance, maybe he just wanted to see who understood the [bleeping] question [grin].)
*Other kind of brain: Spockian.
ADDENDUM: Okay, let's put this another way.
2 3 5 7 9 7 5 3
Eight guys are testing the in-camera resolution of a film. Once the negatives are made, the results can be calibrated scientifically and measured accurately. One guy can't load the film into his camera, but he figures it must be able to form an image or they wouldn't be able to market it; he dispenses with the measurement, wings it, and gives it a 2—the equivalent of getting ten points on a quiz because you wrote your name correctly. Two more guys have poor-quality lenses and can't focus their cameras properly. Their negatives measure as 3's. Two other guys focus properly, but give too much exposure and use the wrong developer for the film because they read somewhere that it gives the "highest acutance." Their results: 5's. Two guys do everything else correctly, but mistake upside-down Peter Max posters for Air Force test targets. Their results are both 7's. The last tester is Ctein. He knows all the potential pitfalls of resolution testing. He's very experienced, having done it many times before. His methodology conforms to widely accepted standards. He uses all the correct ancilliary equipment. He doublechecks all his procedures several times and runs the experiment twice as a control. His result is a 9.
What is the best in-camera resolution the film is capable of, according to the test data?