When an optional AD attribute has been selected for inclusion in AD discovery, if that attribute has a value at time of discovery, it is included in the discovery and reflected as part of the information for the user/computer/group object in CM. If that AD attribute is later cleared (changed to <not set> / NULL value in AD), discovery identifies that the attribute is a NULL value and purposely skips it from the discovery. As such, the user/computer/group object in AD falls out of sync with the AD attributes and continues to report the value even though it has been cleared in AD.

This behavior is clearly documented in all the "ad???dis.log" files, with lines that read:

WARN: ConvertADstoSQLType: pADsValues is NULL
WARN: Type not supported for the follwoing [sic] optional attributes, <attrib1>,<attrib2>,...,

For example. An AD employee has the employeeId attribute populated with their employee number. It is synced into CM. The employee quits and HR clears that information from the employeeId field. It is subsequently not cleared in CM. A similar action can occur with msRTCSIP-PrimaryUserAddress, whereby a once mail-enabled object is no longer mail-enabled, but the CM object still shows the information.

This matches the 2014 Microsoft Connect posting: https://connect.microsoft.com/ConfigurationManagervnext/feedback/details/1000189/pfe-additional-ad-discovery-attributes-when-changed-from-a-value-to-not-set-in-ad-retain-a-value-in-the-configmgr-database