Why do you think a "properly implemented" OpenID site should allow the user to use any authentication provider? The relying party site is trusting the OpenID provider to authenticate its users. Wouldn't sites with real security requirements want to vet providers before trusting them?