Unless I miss something, such a scenario does not seem handled for now.
The process tries to download the centos-release pkg after having disabled all repos from redhat.repo, hence it happens too late in the process.
=> Uninstalling the pkg in question from the RPMDB only (rpm -e --nodeps --justdb centos-release) fixed the issue for the customer.
=> On my side, in using standard/external CentOS repos, I needed to add "7" in /etc/yum/vars/releasever and then hacked the code to remove that file before convert2rhel invokes a repoquery against RHEL7 repos (where $releasever must be "7Server" instead).