The advantage of HMP.A is that it provides some protection against shell code exploits. I am not sure how robust and effective it is -- because testing such things is quite difficult.
NVT ERP can be bypassed by abusing NET Framework and other vulnerable, but white-listed Windows processes. This can be greatly mitigated by adding those processes to the NVT ERP Vulnerable Process list. It's very simple to do...
With AppGuard on the system, those processes will be executed with limited registry and file system access rights.
* * * * *
The whole point to protect the system is not to allow an unknown\untrusted application to execute on the system in the first place. For AVs, if it isn't detected by signature, then is default-allow -- which is absolutely ludicrous and the reason why so many people still get infected using internet security suites.
With AppGuard installed, nothing is going to execute from User Space in Lock Down mode except for processes that are in System Space or on the Guarded Apps list.
So let's say you have a bad browser exploit that abuses NET Framework (vbc.exe, RegAsm.exe, etc) and powershell and manages to get a payload onto the system. The payload itself will not execute (will be blocked by AppGuard). Furthermore, let's say it is "fileless" malware that tries to use powershell, cmd.exe or something else to mess with the registry (like Powerliks) - then AppGuard will block it. It's because all those processes are child processes of the browser - and the browser is a Guarded App with limited rights. Therefore, all child processes inherit its parent's rights = in this case, limited rights.
I suppose if a malc0der really studied AppGuard - and figured out what registry keys it does not protect - they might be able to create some form of bypass. It's not out of the realm of possibility, but not likely either.
However, as it stands, most people that get infected with AppGuard installed managed to do so by making a blunder. For example, not re-enabling AppGuard's protection or blindly installing a program that they didn't verify as safe before executing it.
The general attitude is that security programs should be default-allow for typical users, and yet, still completely protect the system by figuring out for the user what is malicious and what is not malicious. That protection model has been complete bunk for some 20 odd years at this point - and the reason why people still get infected.
The best protected system is one with only very well known programs and one that is completely locked-down (static).
Screwing around -- always trying new files\programs - with most any AV and improper configurations = eventual infection.