The NCVS, for obvious reasons [0], doesn't even track homicide (nor, from looking through the questionnaires it uses, does it appear to gather information on perpetrator race), much less corrobate the conviction stats presented for murder (and, in any case, as victims of crime often rely on law enforcement to identify the perpetrator, wouldn't really be an independent check, even for those kinds of crime it does address, on the UCR for perpetrator demographics, though it can be, for those crimes that both address, for crime incidence and victim demographics.)

[0] "In the last six months, have you been murdered?" is not a particularly viable question.