Apparently the problem with wasting cores/threads (little CPU loading) is caused by the executable assigning all WU to the same cores. The t1 WU use core/thread 0 and the t2 WU us 0 and 1. They are overriding the OS scheduler and assigning the affinity.
On my 64t machine, I'm running 2 BOINC instances with each given 32 threads. Looking at HTOP (linux) it is showing 2 threads at 100% on each CPU so out of 64t only 4 are in use.
Since it has been pointed out to them maybe they will change the executable so it doesn't assign affinity.