15K is too low to be ideal, given the impedance rise of the SLP-05's output to 3.4K at 20Hz. That is less than a 5x ratio. On the other hand, depending on the placement of the sub and speakers, the settings on the sub that you end up with, room acoustics, and certainly the source material, the effects of that will not be noticeable on a lot of material, and may not be bothersome on any material. Assuming, again, that the SLP-05's output is capacitively coupled, and that its rca and xlr outputs are independently buffered.
Also, the 15K value pretty much rules out the possibility of using a y-adapter on either of the SLP-05's outputs, because the combined parallel impedance of the sub and your main amp is only (15 x 25)/(15 + 25) = 9.4K
But I'd start with the power amp to speaker-level input approach first, which (assuming it works as we suspect) avoids all of those issues, and has the additional benefit of providing the same signal to both the sub and the main speakers. REL, in the literature on their subwoofers, makes a point of emphasizing the benefits that can provide in terms of overall coherency, since the sonic effects of the main power amplifier will be reflected in the signals to both the sub and the main speakers.
Good luck!
-- Al
