Very very strange for me

#![allow(dead_code)]

fn main() {
    Struct::new(StructB);
}

struct Struct<T> {
    value: T,
}
trait Trait {}

struct StructB;
impl Trait for StructB {}

struct Wrapper<T: Trait> {
    value: Struct<T>,
}

impl<T: Trait> Struct<T> {
    fn new(value: T) -> Self {
        Self { value }
    }

    // fn a() {
    //     Self::new(StructB);
    // }
}

If you uncomments it, this code would lead an error.
Why ? (Please notice main)

Self is the type being implemented, so it's Struct<T>. The method Struct<T>::new() takes a value of type T, which is a generic parameter, so whoever uses the type can choose it. So it doesn't make sense to call it with a value of type StructB – what if someone calls Struct::<NotStructB>::a(), what should happen, then?

Thanks very much!