Strange string comparison using different locales (C and en_US.utf8)

I compared strings with en_US.utf8 locale and found a strange behavior.

Let's take:

std::string a = "A";
std::string b = "a";

C locale (default one) says that a < b, when utf8 - a > b. However, when

std::string a = "A0";
std::string b = "ad";

Both locales give a result a < b.

Code to check:

#include <iostream>
#include <string>


void cmp_strs(std::locale l, std::string s1, std::string s2) {
  auto &f = std::use_facet<std::collate<char>>(l);

  std::cout << l.name() << ": ";
  std::cout << f.compare(&s1[0], &s1[0] + s1.size(), &s2[0],
                         &s2[0] + s2.size()) << ' ';
  std::cout << "\n";
}

int main() {

    std::cout << "A v a\n";
    std::string a = "A";
    std::string b = "a";
    
    cmp_strs(std::locale("C"), a, b);    
    cmp_strs(std::locale("en_US.utf8"), a, b);

    std::cout << '\n';

    std::cout << "A0 v ad\n";
    a = "A0";
    b = "ad";
    
    cmp_strs(std::locale("C"), a, b);    
    cmp_strs(std::locale("en_US.utf8"), a, b);
    
}

Compare function

What is weird here is that "A0" and "ad" have to be compared by first symbols, A and a, and utf8 returns "less". But when I compare just the first symbols (the first case), it's the "greater".