This bug is closely related to issue 4960 and issue 6745, although I don't think it's an exact duplicate.

One of the simplest test cases I could come up with:
----
struct S(uint N) {
    void f(S!N u) { }
}

alias S!1 A;

void main()
{
    A().f(A());
}
----

Which gives the following error messages:
templatetest.d(10): Error: function templatetest.S!(1).S.f (S!(N) u) is not callable using argument types (S!(1))
templatetest.d(10): Error: cannot implicitly convert expression (S()) of type S!(1) to S!(N)

In this form it is fairly easy to work around, by either changing uint to int in the struct declaration or using S!1u instead of S!1. However, I derived this from a slightly different test case:
----
struct S(int N, T) {
    void f(U)(S!(N, U) u) { }
}

alias S!(1, void) A;

void main()
{
    A().f(A());
}
----

Which gives me these error messages:
templatetest.d(10): Error: template templatetest.S!(1,void).S.f(U) does not match any function template declaration
templatetest.d(10): Error: template templatetest.S!(1,void).S.f(U) cannot deduce template function from argument types !()(S!(1,void))

In this version, changing the type of N or 1 doesn't do anything, and I have not found a way to work around it.