Hello !
First and foremore, duplicut is not meant to be faster than sort -u (but i'm happy to see it is in some cases).

What sort -u does is is sorting the file alphabetically, then iterating through lines to see if line == line+1, and delete if yes.
What duplicut does is actually a lot more complicated, as it is able to remove duplicates without sorting. So if you don't mind keeping the original order, it might be better to use sort, or other tools for duplicate removal.

But anyway, i suppose you used sort -u just to compare outputs and check if duplicut actually works.
So i invite you to run duplicut --help, and take a look at the options.

For example, there is --line-max-size, which defaults to 14, meaning that lines greater than 14 chars are removed, event if unique.

Also, empty lines are automatically removed by duplicut.

These aditional behaviors exist because duplicut is mean to aggregate password wordlists, without losing the order, and without having duplicates. And in a passwords wordlist context, i rarely want to keep lines longer than 14 chars, as they might be a garbage line, a too long password to deserve to be guessed, of a parsing error from the tool that generated this line. Empty line are also deleted for being obviously useless in a wordlist of passwords.

Anyway, if you want to test duplicut, i recommend you to check at these files from my test suite:
https://github.com/nil0x42/duplicut/blob/master/test/scripts/remove-duplicates.py
https://github.com/nil0x42/duplicut/blob/master/test/tests/nonreg.sh

remove-duplicates.py is a small python script meant to behave like duplicut (it's just million times slower :)) so you can read it to see what's different from sort, and you can compare it's output with duplicut's.

Anyway, if i answered your doubts, and if the issue is resolved, feel free to close it. Othersiwe, i'll be happy to debug with you !

Adding needs documentation label, because this issue has probably been caused by --line-max-size option being unclearly documented
Possible fixes:
Add a phrase when duplicut phrase saying exactly how the wordlist is going to be filtered, something like:

removing lines larger than `N` chars, containing non-printable chars, or duplicated.

Another interesting interesting 'user warning' would be to inform user if no \n has been found in file's first 4096 bytes (because file might be an old-style macOS \r newline separated wordlist)

@freeroute , can you please confirm me if the problem was due to --line-max-size option ?