It explains why your numbers don't match.
In a normal exposure, the mechanical shutter controls integration time. There are periods of time where the pixels are "active" (able to collect light), but are being blocked by the mechanical shutter. Mostly this is the time after the shutter closes, but while the pixels are still being read out. FSRP uses basically the same sequence, except the mechanical shutter is not activated. So pixels are receiving light during the periods of time they would normally be blocked by the mechanical shutter.
It takes some finite amount of time to read every line of the sensor, and the lines are read one by one. You can't deactivate or reset the lines before you read them out. So if you don't use a mechanical shutter, you are going to see a brightness gradient, and there's going to be some extra exposure time.
Maybe I can explain better with an example (way over simplified, but makes the point):
t1: global reset (turn the pixels on)
t2: mechanical shutter open
t3: mechanical shutter closed
t4: readout begin
t5: readout finish
Normal photo exposure time = t3 - t2
FSRP exposure time = gradient from (t4 - t1) to (t5 - t1)
FSRP total exposure time* = t5 - t1
*according to a1ex, there is some other stuff happening that is making the reported FSRP time even longer (longer than actual integration time), the time you see on the screen is simply the amount of time the function call to FA_CaptureTestImage took.
