The "right" way to implement an upper limit or saturation really depends on how you want to physically model the effect. What should happen to any reactions that produce this species after it reaches saturation? Should these reactions "turn off"?
As Ingrid said, the most natural way to model saturation is probably using an event. From your comment to her answer, you may be happy creating an event that simply ensures that your species never exceeds the upper limit. However, because of the way that events are detected, this event will not do what you want unless you also "turn off" all reaction that produce it.
It also sounds like your model uses dimensional analysis and/or unit conversion. In that case, here's how you might implement the event:
Let's assume that the species you want to limit is named D, has units of mole/liter, and is in a compartment named C. Let's also assume that D is the product of a reaction with a rate constant K.
First, add a parameter (with units) to the model that sets the saturation limit, for example, DMax = 3.9e-6 mole/liter. Next, add an event with the trigger "D > DMax" and with TWO event functions "C.D = DMax" and "K = 0*K". The first function ensures that D is set back to saturation, while the second function ensure that the reaction producing D is turned off. Note that the trigger must be written using an INEQUALITY not an EQUALITY. Also, you'll ned to set the ConstantValue property of K to false so that you can change its value by an event.
Any more specific suggestions would require knowledge of your model, and as Ingrid said, would best be handled by contacting technical support.