If all authentication is done by typing passwords into the same device (your desktop client) - then, yes, key stroke loggers are still going to be an issue. Keep in mind that when you separate authentication systems, you open up options to go to multifactor authentication, as well - if not now, then in the future - where you could use tokens, separate key pads, etc. - which could limit a simple key logger's ability to get the high privilege access.
Man in the middle - if the man in the middle is in the middle for both levels of authorization, then yes, that's an issue. But most times I'd have been surprised to find that both took the same path. So if you're talking about web browsing, then possibly. But if it's access to the host, then access to something on the network - you may be talking about two paths, and the man in the middle would need access to both.
Social Engineering (including phishing) - Usually multilevel authentication comes with multilple rounds of training. In the case of an admin account and a regular employee account, the employees usually get a long boring training session on the importance of security, which may or may not be effective. Quite often admins with high risk accounts, are given a secondary round of training that is more pointed, and more along the lines of "mess this up in a big, careless way, and you'll be fired". By giving and separating the password, it's often easier, administratively, to enforce this.
Separate security processes - with two levels of passwords, there's an option to let the low level password store compromise on the side of availability while the high risk password store tends towards privacy. For example:
the low level may provide federation services, remote access, services that support non-company devices, etc. The high level store may require physical access to the building, encryption capabilities that aren't available on all devices, and higher-grade password quality.
separate storage - both live and in backup and in logging. A high end system may require encrypted tape backup, securely stored audit logs for a certain period of time - being able to totally separate the infrastructure can be very helpful here. High end storage is expensive, so less to store is a big win.
access monitoring - live monitoring of access attempts - a system far in the front of an infrastructure may invoke so many brute force attempts a day that you don't worry much about it until it becomes a DDOS. A system nested deeply inside the infrastructure may be of serious concern every time the intrusion detection raises a flag. Access monitoring may be staffed differently, monitored differently and reacted to differently.
I could probably dream up a few more.. but most of what I come up with as BIG wins are from a large scale organizational perspective where the focus is not just on the single user and their end machine, but on the system as a whole, and the need to trade cost vs. risk.