The take-home I get from Bangor, Kortun, & Miller (2009) that you cite in your comment is that there’s not much difference at least in the categories they used –all averaged in the 66-76 range, while individual product scores ranged from 30 to 94. With the quartile scores Bangor et al provide, you can convert a score you have into a percentile by making some assumptions of the distribution. Or, to make it easier, use Jeff Sauro's percentile converter and guide for SUS.
I'm not so sure breaking products down into more specific categories will yield a wide range of averages. Form factor (e.g., TV versus phones) influences usability, but I’d expect smaller average differences among functional categories within each form factor (e.g., ecommerce versus social networking on the web). If you wanted a database, the first step would be to define subcategories of products that are correlated with usability. That itself is a pretty big research project.
If you really need to know precisely where your product stands with respect to your competition, then run your own SUS test on the competition using the same user pool, tasks, and testing environment as you run on your own product.