Any coordinate that is effectively zero (and something e-15 *is* effectively zero), as bootchk points out. There are always small rounding and coordinate conversion errors when transformations are applied to floating point numbers. You're seeing one of them.