unicode normalization not implemented correctly

[bagohart]

Here's an MWE, save it as test.jl:

Û = 5
x = Û + 5
print(x)

Using neovim, I input the first Û using U\hat<Tab>. The second Û is input with a unicode picker as "latin capital letter U with circumflex". https://en.wikipedia.org/wiki/%C3%9B
With julia test.jl from the command line, the program works as expected and outputs 10. Looking at it within neovim, the LSP complains that "missing reference: Û".
One of them is wrong, and I guess it's the LSP.
Using https://docs.julialang.org/en/v1/stdlib/Unicode/#Unicode.isequal_normalized both strings "Û" and "Û" are equal, so that is not the problem.