To try to avoid speculation, and to supply a bit of practical testing, I've just swapped my Samsung 1280x768 monitor for an Apple 23" 1920x1200 Cinema Display on this quad G5 PowerMac.
I've just played back the video I previously 'Shared' from iMovie (see several posts above) which now almost completely fills the screen, instead of overflowing it. The quality of the 1888x1062 exported movie looks even better (clearer, sharper, slightly better contrast) than the original footage. But it looks exactly the same as it did on the Samsung, except that I can now see more of it.
I imported the same footage with the 1920x1200 Cinema Display connected, still got the..
.."Selecting Full on this computer may result in degraded video playback" message, though my graphics card's (Nvidia GeForce 6600) display settings are now 1920x1200 instead of 1280x768, and then Shared (exported) it, as before.
There is absolutely
NO difference between the footage imported and exported with the lower pixel count Samsung 1280x768 monitor connected, compared with the footage imported and exported with the Apple 23" 1920x1200 Cinema Display connected.
As mentioned above, it'd be rather odd if the editing and exporting results were of different quality depending on what monitor, or video settings, you had.
Speculation and guesswork doesn't often help very much. If we want to understand the encoding, or decoding, or anything else which goes on inside iMovie, it's really better to actually
test what happens (..e.g; by creating HDV-compatible 16:9 dimensioned two-field frames from, say, Photoshop files of odd (white) and even (black) lines, and then see if iMovie de-interlaces or combines them or discards a field and doubles a field or whatever..) or to actually
measure the dimensions of material imported into, or handled within, iMovie, or to go to the Developer Discussions and to see how the actual
programmer(s) of iMovie have configured its behaviour.
Just guessing about its import and export capabilities or features doesn't really get us anywhere. If we have real, demonstrable, helpful evidence of what it does, and how it actually handles video, then let's say so.
But if we're just surmising, I don't see that it's worth bothering.