Let's start with the definition of a completely secure computer system: a computer locked into a room where nobody has any access to it.
That is also the definition of the most useless computer system in existence.
While far from ideal, in reality there is a very dynamic balance between computer systems that are secure and that are useable.
I retired from the US Navy Reserve half a decade ago. We had a secure network at our disposal which was all that we had to work with there. At the same time, we had to do our administrative work somehow (for my last decade of service, I served mainly as XO, but sometimes also as Officer in Charge (OIC)). What ended up happening was that our admin work was performed on our laptops outside of the Navy and Marine Corps' secure network. The secure network was far too restrictive, as required by security considerations, to allow us to use it for admin work. Yet we still needed to gain access to the reserve center's printers.
Sparing you the details (which you shouldn't know about anyway, for security reasons!), we constantly found ourselves having to find ways to work around the secure network. That shows that in real time there can be a conflict between security and operational needs. Ideally, there should be a secure network that meets operational needs. In reality, ???
... emails containing classified information, ...
Who classifies information? Government agencies do. Do they classify that information before that information comes into existence? No, they classify it afterwards. So then during the original transmissions, what was their classification? Er, none?
Classification always happens after the fact. Was that classified? Not at the time, Ma'am. Afterwards, though. OK, in real time, how do you handle that realistically?