I'm curious as to why you think that the Bureau of Justice Statistics is biased towards undercounting (I'm not saying that it isn't, just wondering why you think it is).
Incidents reported to the police are but a subset of the ones that actually happen, especially when the victim thinks there is nothing to be gained from making a report. While the NCVS does have an advantage over police reports in that it reaches out to survey people, even then people are likely to avoid mentioning incidents simply to avoid the hassle of being stuck in a lengthy poll.
Conversely, the other set includes incidents in which people think they stopped something when there wasn't anything that happened, plus ones that were resolved while having a gun that could have been resolved even without a gun being involved, so it overstates the usefulness of DGU.