Now current solution is clear.. the idea of use activity in rating is good..
I think you could use a unified formula like my solution to have a more accurate rating (giving a weight to activity score, a weight to unregistered voting score and one to registered voting score). It may resolves the "problem" of points overwhelming, because the importance of each aspect has a related percentage. For example you could evaluate popularity in this way:
Popularity = 70 * (ActivityScore / MaxActivityScore) + 30 * (PositiveVotes / TotalVotes)
In ActivityScore you could add +1 for "Website" click and +2 for "Download" click (independently from the solution you will use, I think a Download click is more important that a Website click).
MaxActivityScore is evaluated checking all ActivityScore counters once per day.
For this solution you need also to separate counters of positive and negative votes for each app. But you could consider to add registered and unregistered votes together, giving: +1 to unregistered users, +2 * UserLevel to registered users.
Another good improvement could be to add a simple time factor, for example giving a different weight to scores of different dates (I have something in mind, but I avoid to write a too long message now).
In alternative, keeping current solution for rating, I think these aspects may be improved:
1. give a different weight to Download and Website clicks (as previously described)
2. give a bigger weight to "voting score" (or eventually keep them permanently, not only last 30 days votes)
3. offer the two scores you proposed (last 30 days and all times)