CAPS Zuma (windows 2012, JRiver using Acourate filter in the convolver, ASIO) -> Playback designs MPS-5 -> pair of Krell 750mcx -> speakers
Macmini (with windows 8.1 and REW 5.01 beta20) <-> Roland quad-capture (with loopback in second channel for timing) -> line out to preamp
<- Earthworks M30
Anything else you need to know?
is it right that your measurement setup is completely separate to your playback chain? I don't understand how you're intending to get REW to playback through jriver in this case as REW doesn't have any external signal generator AIUI. To do this, you need REW to be able to route a signal through jriver. I'm not sure what "line out to preamp" refers to exactly but, if you don't have REW on the CAPS box, then you would need that line out to go to a line in on the CAPS and then route that into jriver.
Alternatively if you connected the quad capture to the CAPS then a quick read of the manual
suggests you should be able to use asio line in much like I do. This means you would configure REW to accept input from the mic as usual but output would be to some asio output channel. For the sake of argument lets say you picked output 5. You would then open Asio Line In with channel offset = 3 and open it for 1 channel input. Note jriver uses a 0 based index so if your asio device has channels 1,2,3,4 then jriver sees 0,1,2,3 therefore saying offset = 3 basically tells jriver "read asio channel 4 as channel 1". Assuming this opens ok then open the dsp studio in jriver, turn off any other dsp modules, look at the analyser screen and send a sweep through it. You should see the trace go through the L channel. Once you see this then you can switch convolution back on and away you go.
I am assuming you are only using 2 output channels here hence have 2 spare on the roland. Google says your amps are monoblocks so I guess this is true.
A couple of foibles....
I've noticed that this setup produces 2 channel playback at times so I have to add a PEQ module to mute the L or R channel. I don't know why this is & I haven't investigated it further.
I've never found a way to provide a loopback timing reference in this setup, you would need to route that through convolution too otherwise the discrepancy in the delays between the 2 signals (I think) causes REW to get confused about the measurements (it sees the timing signal, sees no measurement signal due to the delay, errors about low measurement levels). If you need this function then I use holmimpulse & it's time lock feature for that.
If this doesn't give you enough info to get going then I think interact is a better place to ask as there are more people with more experience on this subject there.