Questions to consider when selecting projects to compare against:
What phases of effort have been included?
It is a similar team size?
Is the language similar in terms of effort in usage and expertise?
What methodology has been used for development?
Is the industry a relevant factor in productivity. E.g. Military would certainly be relevant.
How old is the data? Methods might have changed considerably since that time.
Is the duration of the projects similar?
These are some of the filters that you would like to use when checking an estimate. Keep in mind that if you filter too much you may end up with a sample size that is not statistically valid.