Sounds like a bug. Did you try it again (or with another leading-edge model) and get a similar result?