If you replay the tape you will see that the priest was told when he came out of the second "after" pancreatic scan that his diabetes was not quite yet gone and that he had in fact another one percent of fat to lose off his pancreas before his diabetes would be gone. Prof Taylor himself said this, and also that he would need to continue the diet for "about another week"
Ah. Ok.. I'm afraid that's triggered my cynicometer. 1% is often like 97%, especially when it comes to things like certainty or uncertainty. Sounds good until you look at the data, and especially error bars. So for example, here's another study picked mostly because it's the first one Bing found and is well-cited. Plus the author's uploaded it so it's not behind a paywall:-
https://www.researchgate.net/public...ography_according_to_the_duration_of_diabetes
And look at Table 2 on P.11 where it summarises pancreatic volume, fat volume and fat percentage. So their 'Normal' mean pancreas was 66.3cm^3, and pancreatic fat only 1.9cm^3.. But also look at the SD, which shows a pretty wide variation in both pancreatic & fat volume. There are very few normal people, we're all deviants in one way or another..
So losing 1% off their 'normal' pancreas would translate to only around a 0.02cm^3 difference in total fat volume. Which would be difficult to measure accurately. Prof Taylor's technique may rely on being able to more finely slice their patients digitally, or count pixels more accurately. But it's a small change to detect and suggest is the trigger level between being diabetic, or not.
But that's science, and why these studies get done. The Lim study's from 2014 and focused on Koreans, and as they note the variation in pancreas size based on body size and diet, ours may be larger. Not that I'm suggesting Koreans have a small pancreas as I've had a lot of fun in Korea, and their food is awesome.
And there's a challenge in assessing pancreatic fat in general. Prof Taylor has specialist imaging systems to support his studies, most of the NHS does not. I'm also not saying this is bad science, all the data helps refine potential correlations and ways to manage diabetes.