MS Word bookmark and PDF bookmark are completely different entities. In MS Word bookmark is a marked area in the document that can be referenced, read, replaced etc. It has start and end. In PDF bookmark is a point referenced by a TOC entry.
In PDF there are no bookmarks in MS Word meaning. But Aspose.Words implements some kind of mapping of bookmarks in the source documents when converting to PDF. Start nodes of bookmarks are output as text segments with segment IDs. End nodes are completely ignored. If a bookmark occurs at the point where text segment is not acceptable in PDF its output delays until it is acceptable. Normally this doesn’t break anything.
Try converting any document with bookmarks into Aspose.Pdf XML intermediate format. Open as text. You will see the following constructions:
This is what your bookmarks became. After conversion to PDF, say loading to Aspose.Pdf, you can programmatically access these segment identifiers via Aspose.Pdf API.