Creating a DataFrame from Series with unsorted indeces gives completely bad data. I've [written a module](https://gist.github.com/3294509/e777009080fbcbe3c6c91baeb3a71fbb1a092cdc) that reproduces the bug—unless this is by design. Give it a try. (I'm running `pandas==0.8.1`, Python 2.7.2, and OS X Mountain Lion.)