Your maths is a little elementary in this case. You assume it is as likely to capture the flag on any given second when I would argue that it is less likely if allies have just respawned and more likely the more allies are dead. But of course this is difficult to calculate accurately without sufficient data, seeing there are so many variables to consider.
I think you are both right. But as Ande said, to me the best solution would be to just experiment with it. I think you can't really go wrong here with a 2sec adjustment on either side.