The applications are very different, plus the numbers you are crunching are largely different.
SOB would hit about #6 of the top Primes found, WR would be pretty far over the top. 6th spot is over 9 million digits, WR is going to be somewhere over 18 million digits.
So, as you can see there is a huge difference in the numbers. I am not the best at explaining the difference in the programs, but I know GFN does have a lot more optimizations than the LLR does.