Hi Vlad,
First off - I absolutely recommend the book "Web Caching and Replication" by Michael Rabinovich and Oliver Spatscheck if you really want to know the theory and practise of web caching (although out of print you can get second-hand copies easily from amazon, etc). It's not Cisco-specific (doesn't even mention Cisco) but after reading it you'll have a much better idea of how caching works and what's achievable.
If we look at the second question first, (i.e. why is the traffic to the cache lower than the Internet traffic?) then we need to examine what method you are using for directing traffic to the cache. You mentioned transparent proxy (but not whether it was L4-switched or WCCP-directed) and also that some users are using it as a direct proxy? There are a bunch of issues associated with each method (for example WCCP version 1 can only redirect port 80) so it sounds like whatever method you are using simply isn't redirecting all the traffic to the cache. (If running wccp then do a "show ip wccp web-cache" on the router to see the packet ratio of redirects).
If you can force more traffic to the cache then you'll likely increase the savings proportionately - but you may find that a large percentage of the traffic isn't cacheable (e.g. VPN sessions, SSL, streaming media, etc..)
HTH
Andrew.
