Hi,
since no-one else has answered this, I'll try....

you could try using our new backend, mgpp (mg++). This allows
queries to have separate terms for different metadata fields.

We have an online demonstration that uses mgpp for highly structured
bibliographic data:
http://www.nzdl.org/cgi-bin/library?a=p&p=about&c=coltbib&ct=1&qt=1

You can find the collect.cfg file for it (with the others) at
http://greenstone.org/english/docs/collect_cfg.html

Basically, you have a "buildtype mgpp" line in your config file,
and you no longer need to prefix the metadata with document: or
section: in the file. For this example collection, we have a field
that says "text,metadata". This means that an index is created for
both the full text and the metadata fields.

The very bottom line adds a few arguments to the url - ct=1 means
collection type and I think it is needed so that greenstone knows it
is an mgpp collection instead of a normal mg collection. qt=1 I think
means use the advanced query search form instead of the normal, simple
one-line query entry box.

Date searching is a lot harder... I know someone in our group did work
on this a while ago but I don't know how you add it to your collection.

Hope this helps
John McPherson