deannich, DPA does not track buffer cache usage by database. We track at the instance level. So your assumption is correct about the limitation at the database granularity.
Regarding PLE, your algorithm to show percentage of time that PLE is healthy (above 600) seems reasonable. I'd say that going after inefficient queries is always a good endeavor, but at some point, you may just be running up against a resource limitation and need to add memory to the host. I like to think of PLE as an indicator that my system memory is under pressure. From there, you can look at what queries are running at that time causing pages to be aged out quickly.
Overall though, I think your methods are reasonable. 8 )