the min is for minimum cache: whatever memory reclamation is buffers wont go under that value...
So max is to be understood as the maximum buffer cache you let the system have, if there is memory not used that is where it will finish to the value given...
It could release for other programs to do but for that it needs to flush the cache before and that can take some time because lets say it is of 50% of your 24GB that means forcing a 12 GB of immediate write to disk, will / How the SAN cope with that? usually you see in such case a "freeze" of the system ( you know you type and nothing is happening ...) till done... that is why it is resonable to configure that value to what you know you SAN can accept without any trouble ( that means SAN can send an acknowlege almost just after so the buffers can be flushed...)
The system hates to have wasted RAM so if you have unused RAM either it will be used as pseudoswap, or will end as cache unless you dont let it use it ( by reducing e.g dbc_max_pct ...).
If you have JAVA stuff that could run and you are just a little short ( used by cache...) it wont wait for it to be released and exit saying not enough memory...