Since you bound the parameter to be greater than zero, the estimate is indeed greater than zero. Unfortunately, its effect is essentially zero, and if you removed the bound, it may very likely be negative. I would consider reparameterizing the model so that one of two things happens: you don't need a bounds statement, or the parameter b is not included. Since I can imagine that the model is fairly complex, this may be much more difficult to do than to simply say "Parameter b has no effect."
Steve Denham