PDMetadata.exportXMPMetadata fails when Metadata has encrypted stream

When the Metadata is encoded with the Crypt filter, exportMetadata() fails to parse the XML. My guess is that PDDocumentCatalog.getMetadata() gives PDMetadata the raw stream, instead of the filtered one. Then PDMetadata.exportXMPMetadata() calls XMPMetadata.load(), which cannot parse the encrypted stream.
While I cannot post the document (proprietary), the outline shown by PDFDebugger goes like this:
Root:Dictionary(Catalog)
+ AcroForm:Dictionary

• Metadata:Stream(Metadata:XML)
• Filter:Array
  o [0] Crypt
  o Length:6302
  o Subtype:XML
  o Type:Metadata