gps timing query

I've written some code in Vis C++ to display waterfall of wspr sigs etc., and have also input a gps pps pulse on the other channel of the sound card. I know wspr sigs are not timed to us accuracy, but the same code can be used to estimate propagation delay of ibp beacons which are accurately timed. Anyway my query regards the timing experiments done by some wspr people using wwv etc. The problem with doing it my way is that the propagation delay through the receiver is not accounted for, and I know an alternative method involves coupling the pps output to the receiver input directly. Could someone advise the best way of doing this? I've tried an rc type attenuator, but generally get a lot of crud on my display with no obvious correlation with the noise pulses and the pps pulse. I would expect a correctly demodulated pulse artefact to be about 10-20ms after the direct input pps pulse (assuming a standard ssb if filter), but have not seen anything like this yet. Ta.