You would parse the XML using what ever XML tool you prefer for your language you are using. Then you would use our AnnotationsTest sample code to show how to create new annotations based off of your custom ones?
The other option, is to stick with the standard annotations, but add some custom data. Then after importing, parse manually as above, but using the unique id entry, you can match up between the imported PDF annotation, and where you are in the XML, and then modify the annotation in the PDF.