There no real saying in what is a good of bad score.
BAD for sure would be a positive score.
Also you can not compare two different systems. There is no correlation between score and binding affinity for example (we published several paper on this topic).
The question are:
- Is this cluster unique?
- if not, is it significantly better than the next cluster?
- Does it satisfy the data you have?
- Does it explain the biological knowledge you might have on the system?