Like the sibling comment said, this goes against the wording of the TLS specification, but I also think this is looking at the issue from the wrong perspective: from the perspective of the network admin rather than the user. The user does not trust the MITM proxy's fake root. Let's say you set up a corporate network and rather than just whitelisting the external IPs you trust, you give your users the freedom to browse the internet but you pipe everything through a BlueCoat proxy. Your users will take advantage of this freedom to do things like, say, checking their bank balance. When the user connects to the banking website, they will initialize a TLS session, the purpose of which is to keep their communication with their bank confidential. The user will assume their communication is confidential because of the green padlock in their address bar and the bank will assume their communication is confidential because it is happening over TLS. TLS MITM violates these assumptions. If the bank knew that a third party could see the plaintext of the communication, they probably would not allow the connection. If I ran a high-security website, I'd probably look for clues like the X-BlueCoat-Via HTTP header and just drop the connection if I found any.
> As for your list of ways such a system could be circumvented, I don't understand the logic of it. So because there are ways around a security measure, you shouldn't use the security measure at all?
In some cases, yeah. There are a lot of security measures out there that are just implemented to tick some boxes and don't provide much practical value. If they don't provide much value, but they actively interfere with real security measures (for example, by delaying the rollout of TLS 1.3) or they're just another point of failure and additional attack surface (bad proxies can leak confidential data, cf. Cloudflare,) they should be removed. I'll admit most bad guys are incompetent, but it's dangerous to assume they all are, because that gives the competent ones a lot of power, and someone who is competent enough to know that a network uses a TLS MITM proxy will just add another layer of encryption. (Or, like some other comments are suggesting, they'll just test your physical security instead and try to take the data out on a flash drive.)