I think you're not technically wrong, but you're defining NAT differently than the majority of people you're arguing with (those who assume NAT also implies a firewall blocking inbound connections), and the remaining minority (the "on the WAN subnet" crowd) are dismissing outright the idea as a reasonable attack vector that an attacker close enough to be able to send packets destined for non-internet routable addresses to your router.
Is the latter something that was/is actively exploited?
Is the latter something that was/is actively exploited?