In 1979 William Arens published a book asking a plain question about the anthropological literature on cannibalism: what does it actually rest on? He went through the standard cases and reported that in case after case the evidence was second-hand. Somebody had been told; a neighbouring group said so; an early traveller reported it; a colonial officer recorded a rumour. Very little of it was observation, and much of what was presented as ethnographic fact came from people with an interest in the answer.
He also noticed the pattern in who gets accused. The charge is characteristically made about a people next door, or a people being conquered, or a people whose land is wanted. It is very rarely made about oneself.
The stronger claim in the book, that socially approved cannibalism has essentially never been observed anywhere, was attacked immediately and did not survive. Archaeologists in particular had physical evidence, and Mesoamerica is among the places where the evidence is not testimonial at all. Specialists in several regions produced cases that Arens had handled too briskly.
Both halves of that outcome matter. The general thesis is rejected. The source criticism largely stands, and it stands hardest exactly where the evidence is thinnest and the accusation loudest.