Some things that might help:
Regarding seed 2 and the constant 0x5D588B65:
It looks like it's a hash function.
Regarding seed 1:
0x6C078965 used to be part of the old ARNG.
It really looks like this is just another LCRNG with new constants. I used the researcher function in RNG Reporter and plugged in your new constants and got the same results that you did when you forced the seed.
Regarding something I found a few days ago:
Check out the routine at 02019B7A. It's the old LCRNG, but I didn't see where it actually stored the value. Everything came in and out through registers and without a debugger I could not get a handle on where / when it was called.