It's not 10% difference that triggers the alarm, but the 5 mbar pressure rise from the green blinking point, which is 200 mbar below the ambient pressure measured at startup.
Whether the housing is vacuumed or not doesn't really matter, except it changes the housing's depth rating for 2m (if vacuumed to -200 mbar), and that is also more or less unimportant, as nobody makes the housing which will crash at exactly that depth. The only thing that matters is a pressure difefrence, so a vacuumed housing will function exactly the same as a non-vacuumed 2 meters deeper. So, the Buddy dive's argumentation makes no sense, as the o-rings will behave exactly the same, it's just a question of depth. All housings are designed to function with higher pressure on the outside, and the vacuum only helps keep doors and domes tight in place.
How much water must enter the housing before the alarm is triggered depends on how many pump strokes are pulled after the LED starts blinking green. As V3 doesn't have the temperature compensation implemented, it is a kind of tradeoff between the amount of water and a possibility of false alarm due to temperature change, but generally, one stroke should suffice to avoid false alarm in average sized housing, and to keep the water amount at bay. 6ml of water per liter of the housing volume is not an enormous quantity of water, and I think it's hardly enough to trigger any leak detector. Any sanitary pad will handle much more, and I'd suggest using one no matter if you have LS or not.
V4 solves all the problems above, as it measures the pressure when the pumping stops, and takes that as a reference, so it doesn't really matter how many strokes you pull after the green LED starts blinking (but don't go overboard for obvious reasons).
Of course you should operate the buttons and controls during the test, I never put that in the manual as that seemed a no brainer to me, LS is just an electronic replacement for a good ol' bucket;) And that was obviously the cause for the problem tc_rain had, the housing was sealed until the lever was operated...