The way a 2D constraint solver works is that it uses the initial conditions to determine the final shape. So my guess is that, given your constraint system, .5 is a good initial condition and "others" are not good initial conditions for getting an answer. So its not that you are editing it twice, it's that you are editing it off of .5 as the initial condition.
Would'nt' be able to say why this is, but its likely something unique in your constraint system. Probably something to do with what appears to be a equal constraint in the upper right, but that is speculation without having the model.