JG-
I ended up getting aggressive with the value on the parameter 'unknown endpoints clean up' interval setting the value to 7 days. I had a number of stale devices with no attribute including the one you mentioned. Many were gone the next day.
Also I discovered that the parameters 'known endpoint clean up' interval and 'profiled unknown endpoints clean up' interval will not accept a value over 180.
These values are located at Cluster Wide Parameters>Clean up Intervals tab
I too struggled with the documentation. We are fortunate to have a test bed so I interpreted the documentation and tested/monitored in our TB in order to get the correct combination for our environment/configuration. I am deploying in production next Tuesday.
Hope this helps.
KI