Ok, I think I'm being nitty here, but maybe these pictures will help.
I gathered that your agument said lean towards a call in those close decisions because assuming a flat or normal distribution around your estimation of how often you'll win, it's possible that the actual likelyhood of winning is a lot higher than you think, but can only be a little lower because it can't be lower than zero. (That was a mouthful)
What I wanted to consider is that you can't assume an even normal distribution around your guess.
I made these two graphs by freehand in Paint, so be nice.
In graph A, you think need to be good 5% to call profitably. Since the distribution is even, there's more area under the curve to the right since the left side is cut off. Thus, you're more likely to be better off more often than worse off. This would lead towards a call.
In graph B, the distribution isn't normal. Notice that the zero cutoff on the left has caused the bell curve to 'bunch' on that side. Thus, the areas under the curve on each side are fairly close and thus you can't necessarily lean one way or the other. I would think that graph B is a more accurate representation of the way things are when dealing with small percentages.
Did that make any sense and did it have anything at all to do with what you were saying? I think I'm just making things more confusing.