The answer you received from @fdonnelly is on the dot. ESRI uses their own metadata template, which comprise metadata elements from several metadata standards (FGDC, ISO, Inspire, etc). The idea is to provide a catch-all-needs style, which has two key features:
- it is the only one that allows metadata editing.
- it assembles all elements from all the supported standards. This shows well in the graphics provided by ESRI in their help page about metadata formats:
You can always export your metadata into a standard format (for instance ISO19139), effectively filtering it out to the contents of the standard you need.... but the exported version will not be editable any longer... but you do retain the original, ArcGIS-style, editable version.
See it as your work copy, continuously updated as you data change; whereas the export is your well-formatted deliverable.
Our office looked into the mecanism under the hood - the metadata is encoded differently between the ArcGIS metadata style and the standards styles (like ISO19139).
Direct consequence: the XML file you can generate will have a different structure, notably different tag names and organisation, and different content (ArcGIS-style XML will contain more than a ISO19139 XML). This is easy, and (but) does not explain the fact that standard metadata is read-only in ArcGIS.
As we were told by our local ESRI rep, the ArcCatalog metadata editing interface is tightly bound to the ArcGIS-style structure, and is therefore limiting when you want to work with other metadata styles or standards: it is (probably) easier for ESRI to design a conversion process, using a series of XSL files, between the ArcGIS metadata file and its ISO19139 equivalent, than to recode their interface for the various metadata standards they support.