Well, for the moment the Splitters isn't managing to generate a surplus of work, so filling-up cache is difficult. But, there's some reasons that can hinder filling-up a cache of work even then it's lots available.... Some of the reasons are:
"Cache an additional N days" only works with v5.10.xx and later BOINC-clients.
Computer is using wrong Venue, so is using another sets of preferences than you expect.
You're using an override-file with too-small settings.
If you get a wu with short deadline and has a large "connect every N days"-setting", work-request can be blocked until queue empty or idle cpu (on multi-cpu). Not sure if same is true with large "additional-days"-setting...
Work for other projects is in deadline-trouble, or, for some reason you "own" other projects cpu-resources...
Check the computer's info, near bottom there's 4 fields:
% of time BOINC client is running
While BOINC running, % of time work is allowed
Average CPU efficiency
Result duration correction factor
The 2 fields with % shold normally be close to 100%, while "Average CPU efficiency should be close to 1. If any of these is significantly less than 100% or 1, it means you'll get little work each time asks for work.
"Result duration correction factor" will for SETI@home normally be below 1, close to 0.1 won't be a problem. But, if it's close to 2 or maybe even higher, something is out of whack...