DOCBOOK: customised docbook

This is perhaps a non-programming issue but we would appreciate some help
pointing us in the right direction.  We are attempting to convert about 800
books presently in Quark into XML using the 'docbookx.dtd' beta 4 or 5.
>From the exhaustive list of elements, our content (text and images) probably
only refers to main items like title, author, section heading, paragraph,
image, etc.  How do we go about creating a working model of a DTD for our
purposes that will still adhere to the 'docbookx.dtd' specifications?  We
can then use tools such as Adept by ArborText or XMetal from SoftQuad.
Please advise.  Thank you very much.
Neil De Odhar
Data-Flo, New York
e-mail: 
web: http://www.data-flo.net