I would question the logic of using a tolerance that - XXX to -XXX. Appears to add nothing but confusion.
This convention evolves from the use of shaft basic and the hole basic systems. Even under the Geometric Dimensioning & Tolerancing, if a 20 mm diameter shaft has to assemble with a 20 mm diameter hole, tolerances on size, location and form will lead to worst case envelopes (Virtual Conditions... as defined by the GD&T standard) on the hole as well the shaft. These virtual condition envelopes will have to clear each other if assembly is desired. The 20 mm nominal diameter hole will then have both the tolerances on the positive side, and the 20 mm nominal diameter shaft will have both the tolerances on the negative side.
This could be one of the explainations. For process control however, as Martin suggests, there is nothing special about it. You have to put your process center on the mean tolerance.