I can confirm the issue on my machine. May be a bug, or may simply be a product of encodings.
It appears to be the case that your "naïve" (as transmitted to me through
elfeed) uses a regular "i" with some sort of visual diacritics rather than the unicode "ï". You can tell by putting the cursor on your "ï" and running
M-x what-cursor-position, which should return something like:
Char: i (105, #o151, #x69) point=321 of 343 (93%) column=1
Here's what you get with "naïve" when you use the unicode "ï":
Char: ï (239, #o357, #xef, file #xC3 #xAF) point=446 of 449 (99%) column=61
If you use unicode,
evil will realize that you're talking about a single character. So: unicode encoding and entry in probably the most effective way to fix the problem.
The practical step if you can't do that is to move one more point over. So, for example,
vll visually highlights "naï" but yanking only gets you "nai".
vlll visually highlights "naïv" and yanking gets you "naï".
(Note that I've mixed-and-matched the unicode "ï" in here so that the visuals will show up.)