The video noise thing they probably didn't bother to explain because they figured it would take too long. You'll see when I try to explain it below.
For the too high of a brightness, look at the delta tracking graph. The x-axis is the brightness setting of the monitor, while the y-axis is the deviation of the monitor's actual color from the requested color (the higher the number, the greater the deviation is). Note that it jumps all over the place until you make the brightness 70 or above (out of 100), which means that you're only getting good color fidelity at the higher brightnesses -- which is what Tom's Hardware meant. Supposedly, the "ideal" brightness is at about 110 nits, which for this monitor would be at x = 44 (on the graph) -- far above that and it gets too bright. On the other hand, I'm confused by what their "good color fidelity" means. A delta of 3 is supposed to mean "noticeably different to the average joe schmo" (bad), a delta of 2 is supposed to mean "hard to tell unless you do a lot of photo stuff" (good), a delta of 1 is supposed to mean "we can measure this only because our colorimeter is way better than the eye at noticing slight shade differences" (really excellent to just about perfect). For the 2405, it stayed at 1 or below from a brightness setting of about 8 on up, yet apparently a delta of 1 wasn't good enough for Tom's Hardware. On the other hand, note the target audience: professional photo guys. So they really have reason to be stingy about the colors. And considering how much they (presumably) paid for the monitor, they were probably expecting perfection, and were disappointed when it fell short of that. I would imagine that the rest of us don't care once the delta gets below like 2 or 1.5 (I probably wouldn't care once it got below 3, knowing me).
So generally, unless you're a photo pro, the colors are likely good enough; note that the delta stayed at one or below for all but the darkest brightness settings. And the response time is pretty good as well. Remember that a regular 60 Hz frame is 16.7 ms long, so the 23 max (around 20 ms average) is pretty good. You should go to the store and see it for yourself though, because yours truly accepts pretty much anything below 30 ms while gamers demand 8 ms or less.
As for video noise. As panel sizes get larger, the electrical characteristics of the panel itself (as opposed to signal quality, i.e. VGA vs DVI) become more and more of an issue. That is to say, the fact that the little wires have a non-zero resistance, a non-zero capacitance, draw voltage, etc., and thus interfere with each other's signals, become more and more of an issue. This has nothing to do with VGA or DVI connection, which only describes the interface between the computer and monitor -- this is about the architecture of the monitor itself. Thus, video noise. This is why Tom's Hardware uses a checkerboard pattern (the ANSI standard) to measure the contrast ratio, rather than the manufacturer's way of the brightness of an all-white screen divided by that of an all-black screen.
That's the short explanation. The long explanation is this. The panel consists of scan lines (horizontal) and data lines (vertical). So a regular 17" monitor (1280 x 1024) has 1024 scan lines and 3840 ( = 1280 * 3 since each sub-pixel has its own) data lines. Most of the time, the scan line is off (open). When it comes time to render that particular line's colors, though, the line connects, then all 3840 data lines tell their respective sub-pixels what shade to be. Then the line closes again, so that the next line's pixels can be rendered.
Now consider your regular 60 Hz signal. All that data has to be sent then, during each frame. There is also another about 30% of the time that the hardware needs, so the actual data is actually sent only during 70% of that time. So this means that for a 17", the "open" time per scan line is:
1/60/1024*.7 = 1.14 * 10^-5 s
That's right, each line has only about 0.01 ms to transmit the data. By the way, this is why sub-pixels are arrange in rows (i.e. RGB left to right) rather than in columns (RGB top to bottom). The 3840 data lines (vertical) send their info simultaneously, while the scan lines (horizontal) have to wait their turn in line, and when it's their turn they don't got much time to be open. The problem is even more acute for 1200 vertical resolution monitors -- they have even less time to send the current on to the sub-pixels.
Now another problem comes up with increasing the resolution of the panel. Each sub-pixel draws a bit of current, and that means less for the next sub-pixel down the line. So basically the "open" signal gets a bit weaker and weaker every sub-pixel it encounters. Also, increasing the resolution means increasing the size, and the longer each horizontal (and vertical) line is, the more resistance there is in the wire. That sucker has 1920*3 = 5760 sub-pixels per scan line, folks. A thinner wire (to increase the brightness and reduce the "screen door" effect) makes this problem bigger. When you got tolerances as small as this, mistakes happen. Sub-pixels have capacitors, and they may fire when they shouldn't, based on seeing a really degraded signal. Or they may not fire when they should because the signal's too weak. Either way, this ends up creating video noise when you watch lots of rapidly-changing images, where the signals are rapidly changing.
Or maybe the Dell 2405 is just so good that the inherently video quality of the movies chosen by Tom's Hardware is finally noticeable.
