If the data were collected using GPS and haven't been massaged after post-processing those features can often have 'loops' in them where the GPS has vertices that are captured in relatively close order but are mal-positioned, or might even reflect the path that the field person took while collecting the data. Another 'weird' thing that can happen with gps data (though that is not this case) is that if areas are collected by traversing a boundary in opposite directions - say, where you walk counterclockwise around one area, and then clockwise around another. Its odd and doesn't show up wrong or every time but this is often the problem when do things like calculate areas and start getting negative values or hatch symbols that display 'outside' the features. GPS is a wonderful tool, but the data can bring a whole new range of management issues.