When testing on myself (only 1 data point, I know) it actually seemed that having the first letter in the correct position made the biggest difference. So even on long words having first and last letters placed correctly made the biggest difference in readability for me compared to words that had more than 2 letters in the correct position but in the middle.

From playing around with this at worst they were equally unreadable[0], the best case was the first/last letter helped. Eventually I'd be able to decipher any word but by no means did my brain just visually did the replacement "on the fly" or close to it.

It would be interesting to see the study debunking this and the methodology. Do they take into account how many position each letter is shifted, how many had switched order not just shifted to the same direction or same number of shifts, will phonetic languages make a difference (in English removing one letter can vastly change the phonetics and make you miss the similarities with the original), etc.

[0] A quick and likely imperfect example resulting in words that cannot be recognized despite the high commonality:

something [original]

stnhioemg [1st/last]

msgethion [4th-7th]

gnmethios [3rd-7th]

gnihtemos [reverse spelling]

I will debunk your debunk.