This was my takeaway as well.

The pdf says there's a 95% accept rate on their takedown requests. They use that as evidence of censorship but, to me, that looks like evidence of judicious requests that meta agrees with.

Without data on what was taken down, there's no way to explain the difference. There's no reason not to make the entire dataset public (anonymized if you'd like but, since the content is implied to be benign, what's the harm in not?) and show some examples.

The implication that, because Israel submits the most requests that they must be acting in bad faith makes sense only of all countries had an equal amount of content generated that they'd like filtered. It's very easy for me to believe that Israel would have more content directed towards it that violated the Meta TOS.