Facebook removed 95% of hate speech identified by automation / 5% manually of the hate speech they could identify which was between 3 and 5%.

Facebook also said they removed 99% of terrorist/isis posting. Of which the author claims the total amount is unknown. But we know the % of hate speech because of facebooks ability to guess (as stated by the author of which the author was confident). Now we are lead to believe that terrorist content can't be estimated by facebook.

Either we accept they can estimate and take the 99% of terrorist content removal in total at face value (of the total amount) or we must ignore facebooks hate content estimates