The question I'm going to pose is surely, in some extent, version dependent, however I conjecture that it's rooted in general aspects of video manager architecture in Windows. This, I believe, makes it of overall interest.
Taking a screenshoot of the whole desktop by means of:
"Needs["GUIKit`"]; fsa = "FullScreenArea" /.Flatten[ SystemInformation["Devices", ScreenInformation"] ]; saveScreen [fsa_, fileName_] := Export[ fileName, GUIScreenShot[fsa], "BMP" ]; saveScreen [fsa, "e:\\screenshoot.bmp"];"
I get this picture
conversely using a screen grabber (Wisdom Screenhunter free) I obtain this one:
The difference is presumably due by a screen dimming program running on my machine (DimScreen v1.1) and could be easily suppressed setting the dimming factor to zero.
But this isn't the main point; rather, what does this "symptom" uncover (if any) about the way Mathematica reads the video memory ?
May be, pops up something usefully employable ...
( Mathematica 126.96.36.199 , Ms Windows Xp 5.1.2600 )