This is one of those infuriating "it's always been like this" bugs with the monitor.
The monitor lazily initializes itself with information from the Master. One of these is the Instance Name for the cluster. So, when you first visit the monitor, you'll notice that you get "Instance Name: null" in the header.
Eventually the actual instance name will get filled in. We can do better than propagate "null" to our users. The monitor knows the directory in HDFS and thus the instance ID. It can also get the instance name from the mapping in HDFS. Best as I understand, it's just lazy that we don't do this.