Those portion size group cutoffs reek of fiddling definitions until they could generate a result from random noise. They're not doing anything relative like comparing 10/90th percentile processed meat eaters. And even for the "high" processed meat eaters they're considering people who are still eating very little processed meat, so there's lots of room for minor biases from imperfectly measuring and controlling for the effect of other causative factors to significantly skew the results.

Those portion size group cutoffs reek of fiddling definitions until they could generate a result from random noise. They're not doing anything relative like comparing 10/90th percentile processed meat eaters. And even for the "high" processed meat eaters they're considering people who are still eating very little processed meat, so there's lots of room minor biases from imperfectly measuring and controlling for the effect of other causative factors to significantly skew the results.