I have a book I’d like to convert from Markdown to epub via Pandoc. The book is all plain text and does not contain any figures, formulas, or code. I need it to consist of several unnumbered parts with each part divided into chapters; the chapter numbering does not renew with each new part. So something like:

Book Title
 - First Part
   - 1. First chapter title
   - 2. Second chapter title
 - Second Part
   - 3. Third chapter title
   - 4. Fourth chapter title
   - 5. Fifth chapter title
 - Third Part
   - 6. Sixth chapter title 

With all chapters in separate files, I have something like this for part 1:

# First Part {-}

for the first file in which just the part heading is specified, and then

## First chapter title

Lorem ipsum dolor etc etc

for all the chapters (that’s a setup for one directory in which the part is contained, and I have several of these directories). And I also have a title file in the root.

The command pandoc -o draft.epub title.txt part1/*.md --table-of-contents --number-sections --top-level-division=part outputs a table of contents that looks like this:

Book Title
 - First Part
   - 0.1 First chapter title
   - 0.2 Second chapter title

Which is not what I need, it has these weird zeroes. But if I try to shift the heading level with --shift-heading-level-by=-1, then the part entry First Part takes over the place of the title, so it’s now like

First Part
 - 1. First chapter title
 - 2. Second chapter title 

That’s not what I need either.

How can I achieve the desired effect? Top-level headers (#) as parts, second-level headers (##) as chapter titles, parts are unnumbered but present in the ToC, and chapters are numbered like 1, 2, etc and not 0.1, 0.2, so on.