Sorry, I didn't notice his hyperbole. However, I, too, would like to see the numbers of functionally illiterate who graduate high school. My impression is that the inner city schools (black and hispanic) are far more likely to graduate students who can't read. The white illiterates tend to be drop-outs, in my experience.
The hyperbolic point you are arguing against may be damning the public school system for failing the black graduates, not claiming a genetic component for the failure.
Words like "only," "all," and the like generally paint one into a statistically untenable corner. Only need one counterexample to show the statement to be false. Which is why I avoid them. OTOH, such use of an absolute may be true for the writer/speaker if the sample of crimes he has witnessed were such that he only saw the one sort of perpetrator. For instance, we had a business owner locally whose business and home was in a rough neighborhood, mostly hispanic & black. But, it turns out that the only tweakers he had to kill with his shotgun were white. Three of them in two weeks before the word got out and tweakers stopped busting into his home/business. For him, the "only" criminals he had to kill were "all" white trash.
Every once in a while you see a quantifiable situation with a large sample size where such absolutes might be used, depending on how one wants to round the numbers and speak somewhat sloppily. For example, "In 2008 the 'only' black & white interracial rapes were between black rapists and white victims." (
http://www.bjs.gov/content/pub/pdf/cvus/current/cv0842.pdf Table 42 Percent distribution of single-offender victimizations, by type of crime, race of victim, and perceived race of offender)
The table shows 0.0%* for white on black rapes, showing fed.gov rounds to the nearest tenth of a percent. If you read the "*" note, it indicates 10 or fewer white on black rapes reported. Doing the math shows that the proportion to be from 0.0000117 or 0.00117% (assumed # victims =1) to 0.000117 or 0.0117% (assumed # victims =10). Of course, if "10 or fewer" happens to actually be zero, the percentage is zero and an absolute sort of term would be factual. Frankly, rounding to the nearest tenth of a percent is legitmate in this case
Doing a bit more math, there were 19293 black on white rapes in 2008. Assuming the largest number of white on black rapes indicated (10), that shows that there were 1929 times more black on white than white on black rapes in 2008. Or 19293x if we assume the number of white on black rapes to be one. Or infinitely more if we assume the number of white on black rapes to be zero.
The "0.0%*" of white on black rapes has been replicated over several years until recently...
when fed.gov stopped giving out the numbers under the Obama administration. "If thy number offends thee and causes thee discomfort, pluck it out."
As with any social science data, there are some squishy factors that can be bandied about. In lots of data tables, fed.gov conflates hispanic and white data, driving up white crime rates. The offender's race is that perceived by the victim, not some sort of genetic test of "blue ribbon race determination panel" like they had in South Africa. (I don't really mind that we don't have such panels, as they are creepifying.)
FTR, most violent black on white / white on black crime rate differences are not the ~2000x range. Most are in the 40x to 8x range, depending on the crime. For some others (especially white collar crimes), the rates are flipped and whites commit more.