Eric -- right. When I've worked with it, I've had help, too, and my time learning the most about practical PCR methods was also the time I was getting sick, so it's all a little blurry for me. I would assume that they'd be using very solid software that would specifically exclude primer dimers where feasible and would find a really appropriate match for good agreement of annealing temps, but testing that all out in reality as opposed to theoretically is still a really important aspect of the design, so far as I'm aware. Especially with a brand new nested PCR assay, I could see the possibility that one or the other set of primers could go awry even with all of the genome carefully considered and the temperatures matched. Software and real-time and so forth are making PCR easier, but I suspect we both know from lab experience that there are lots of ways for specific issues to be really finicky in unexpected ways sometimes. Figuring those out and correcting for them can be a real hassle.
I haven't honestly looked in a while, but I feel like I recall that some of the PCR was aiming for parts of env, which if I remember right can also be fairly variable, or at least it is in HIV. If there's some variation in the primer binding site, that could also be an issue, though they should be able to find that out easily if they sequenced the PCR- positives.
I'm sure they've considered all of these things with much more skill and knowledge than I have, but I'm not sure to what extent they've found them to be problems or insignificant or been able to correct for any that might reduce sensitivity. If they were able to culture from samples that had been PCR-, though, I'd guess that there's a kink somewhere damaging PCR sensitivity that they still need to figure out. If the other positives were all antibody, then the stages of infection and the chances that active or latent virus is hiding out in CSF or someplace should be looked at to improve the overall diagnostic sensitivity, which is really true regardless and I'm sure is probably already on the "figure this out" list.
I also can't recall how many samples were sequenced, either, but I know they did some sequencing, and I'm far less worried about specificity with the nested PCR -- they definitely found what they were looking for strongly enough that I can't argue with the positives in the research, unless there was contamination or somesuch. The negatives later corrected do trouble me, and I'm sure they trouble them, too. Finding something in research where you've got the means to prove it in multiple ways is just different from finding it in patients, too, so though I think the research results are solid, I'm dubious about the diagnostic use of the tests until they're getting closer to the same answers via different methods.