Most floating point numbers cannot be represented by a native float. You're suffering from rounding errors.