Some of the MIP models in http://www.amazon.com/gp/product/0471150436 (these models are written in GAMS) seem to be missing OPTION OPTCR=0, so some non-optimal solutions are presented. By default GAMS uses a 10% relative gap criterion.

E.g.:

But I see:

Next month teaching a GAMS course in Germany. I’ll be sure to mention this (and how to spot this in the listing file).

GAMS is quite unique in having a default with nonzero gap optimality criteria. Other systems typically like to solve to optimality unless instructed otherwise.

PS. See comment for another system with a nonzero gap tolerance.