Huh. Normally Mobiunpack creates one with the same filename as the extracted html + .opf extension if the info is available, but maybe the book you're trying it on just has poor metadata.
ETA 2: I think your problem may be linked to the version of the script you could be using. Don't use the original version posted in the OP, but scroll down to pdurrant's modified version in post #5. That's the one that outputs the .opf and any pre-existing images, which you can then swap out.
In any case, you can just copy/paste this to give yourself a rudimentary OPF to point KindleGen at:
Code:
<?xml version="1.0" encoding="utf-8"?>
<package unique-identifier="uid">
<metadata>
<dc-metadata xmlns:dc="http://purl.org/metadata/dublin_core" xmlns:oebpackage="http://openebook.org/namespaces/oeb-package/1.0/">
<dc:Title>BOOK TITLE HERE</dc:Title>
<dc:Language>en</dc:Language>
<dc:Identifier id="uid">JUST MAKE UP A NUMBER</dc:Identifier>
<dc:Creator>AUTHOR LASTNAME, FIRSTAME; COAUTHOR LAST, FIRST</dc:Creator>
</dc-metadata>
<x-metadata>
<output encoding="utf-8">
</output>
<EmbeddedCover>RELATIVE FILE LOCATION OF YOUR IMAGE/IMAGE.JPEG</EmbeddedCover>
</x-metadata>
</metadata>
<manifest>
<item id="item1" media-type="text/x-oeb1-document" href="LOCATIONOFYOUR.HTML"></item>
</manifest>
<spine>
<itemref idref="item1"/>
</spine>
<tours>
</tours>
<guide>
<reference title="Table of Contents" type="toc" href="LOCATIONOFYOUR.HTML#filepos740" />
<reference title="Start Here" type="start" href="LOCATIONOFYOUR.HTML#filepos2751" />
</guide>
</package>
This is directly extracted from a Mobi I just made from a cleaned-up Fictionwise book I reassembled.
For the #filepos bits, you'll have to dig into your extracted HTML and find the corresponding numbers for the parts just in front of the Table of Contents (if it exists) and where you want the book to start (dedication, prologue, etc.).
If there aren't any such places, just edit out those <reference> items entirely.
Then do kindlegen -c2 NAMEOF.OPF -o NAMEOFDESIRED.MOBI. You may want to add a -verbose and a -unicode for both troubleshooting and extra character support. -c2 gives you the tightest compression, though to cut down on bloat you may also want to run the
KindleStrip script on it afterwards.
Hope this helps.
ETA: There's always the option of using
MobiPerl, which apparently has a handy
GUI version for Windows users, too.
It can do a lot of things faster and easier and less intrusively to the output file than Calibre. I often use it to fix metadata on existing files I don't want to otherwise tweak.
ETA 3: And it just occurred to me that you could probably use it to add/change the cover image and nothing else without having to go through all this dissasembly/reassembly stuff.