The problem you are having is drift. Drift occurs between the two sources as a result of the difference between the two clocks. While it should be very close coming from two JB3s, it would not be unusual to have as much as 50-100 milisecs of drift over a long show. To correct drift, you need to time-scale one of the sources to the other, which will serve as your master clock. (Time-scaling can also be referred to as changing 'pitch'.)
I find chopping up one source and pasting it in aligned to the beginning of each track on the other to be unduly cumbersome. And, it is not an accurate way to make a matrix because you will still have drift between the sources which may be noticeable by the end of each track, especially if it is a long one. So, you may end up with things perfectly aligned at the beginning of a track, but it will sound echo-ey or chorus-ey by the end. It is much better to put the two sources next to each other in your edit window and then time-scale one to the other.
To do so, find a loud transient sound (snare hit, mic pop, etc.) as close to the beginning as possible and line them up. Then, go to the end and find a similar loud transient as close to the end as possible and see how far off the two sources are on your time ruler. Calculate the amount of drift and use whatever time-scaling function is available in Wavelab to either stretch the shorter source or shrink the longer source to get them to line up over the entire show. In some software, you grab then end of one source and drag it using the time ruler in the edit window to time-scale by the proper amount (this is how AudioDesk & Digital Performer do it); in other software, you calculate the percentage of the difference between the source in original form and the properly time-scaled outcome you desire: i.e., shrinking a source by 99.999975 % or something like that to make the longer source match with the shorter one. Generally, your editing software will create a new file/track for the time-scaled version and it will take some processing time for it to complete before you may then get your mix together to go from 4 -> 2 tracks.
I do not know Wavelab and whether or not it has a time-scale function, but I would be surprised if it did not. Once you mix two sources from different clocks a few times, correcting drift gets a lot easier and should become something you may consider a relatively easy step to make a matrix from two separate sources. Now, I tend to agonize more over setting the levels of the two sources and whether or not to spot mute any artifacts out of the mix if they are bad and only present in one source.
Maybe someone who knows about time-scaling in Wavelab will chime in. Feel free to PM me if you get stuck.
By the way, you should use the V3 source as the 'master clock' assuming you patched in the JB3 digitally. The V3's clock will be far superior and more accurate than the internal clock in the JB3.