It doesn't matter is it behind haproxy or not. The postfix configuration will be exactly the same.
The exact choice of mail exchanger hostname is relatively free, what's important that it should be the publicly resolvable hostname (e.g. in the domain you registered, e.g. mymailserver.example.org, certainly not end with "local" or anything else like this). This hostname should in general should be specified in the MX record for the domain, it should resolve (for external servers) into external IP (which eventually leads to haproxy). The reverse record (PTR) for that IP should lead back to this hostname.
When Postfix exits into internet it also should use this same IP, so other servers will see things coherently: connection made from IP which points to the name, which leads back to this IP and which the server also specifies in its EHLO. This will greatly increase your chances to not to be rejected as a misconfigured server or as a spammer.
This hostname should be configured as myhostname in the Postfix's main.cf
I don't see much benefit in using haproxy in front of public mail exchangers. Postfix is very secure and robust to be on the bastion frontline itself, it doesn't need an external protection on the level haproxy can offer. It's much easier to make a service highly available for external servers by simply having two (or more) different mail exchangers and MX records for the domain (and that will certainly will be more robust than any setup you can create with haproxy, simply because it requires less moving parts).
It could however be reasonable to load balance the submission service for your own clients. Anyway if you reached the scale when you really need this, you will have a bunch of mail servers, split them by function (some machines do reception from external, some run submission service, some run IMAP service, some web service and so on) and in this case to suggest a concrete configuration we need to know which kind of a server you are going to set up.
