The problem with the data they provide is it is not relational, missing huge amounts of movies, leaving you with no way to distinguish between movies other than title which gives you mismatches when movies have the same name in the same year.
Yes, imdb is a pretty poor source of data. From my research, a combination of the well organized freebase film database[1] (~19mm facts) with details filled in from imdb is a better approach. However, processing data from freebase is not trivial and requires a decent amount of time investment to grok.
Corelate and infer from the data. If you can't distinguish between similarly named films, you still have the director and cast and years when they are active, along with country, production studio and countless other attributes. A single data source is never effective anyway.
Along with data sources you will also need a production process and streamlined workflow. It's a holistic, iterative exercise.
I have a Ruby project that parses IMDb data into a database, and which can also supplement itself with data from Rotten Tomatoes and Freebase. It's fairly generic and could be extended. Email me if you want to chat.
