Just to be awkward, I couldn't disagree more with this idea. Of course no-one agrees on what cars are 'important' enough to remain in the database: different people demand different levels of detail from a resource like this. The only certain way forward is to include literally every piece of data you find, as the more detail you include, the more interesting paths you can navigate through the information. For example, you could find out whether the breakdown of cars spotted in films reflects that of the time, or if the choice of vehicles has been somewhat 'stylised'. If you isolate all the cars used in back-projections (say, in Breakfast At Tiffany's or any number of Rock Hudson films), you could even find out whether some footage reappears from film to film. Deleting cars would be a fool's errand simply because you don't know what uses the database could be put to in future.

A neutral resource, such as a database, loses all worth once an editorial hand is introduced to prune its contents. It would allow all sorts of agendas and vested interests to bubble up, and forfeit its reputation as an exhaustive and impartial list. After all, isn't that the point of the star-rating? Of course there will be a disproportionate number of one-star cars, as that end of the list is more given to reporting the historical kerbscape. As a consequence, it's possibly more interesting than the Chitty Chitty Bang Bang end of things.

If the database is beginning to take up too much server space, how about buying more and establishing a paid membership, in the manner of Movie Mistakes? Their membership costs ten American dollars per year, and members are rewarded with all sorts of perks. Maybe a similar system could work here?