Not All Headers Are Identified during PDF to MD conversion

### Bug
When converting from PDF to Markdown, not all headers are being identified. For example, for the attached PDF, the headers Sec. 28-4. - Applicability. and Sec. 28-5. - Repeal. are identified correctly, but Sec. 28-1. - Short title. and Sec. 28-2. - Purpose., etc. are not identified as headers.

### Steps to reproduce
Use DocumentConverter() to convert pdf to md ([Stafford County - VA Zoning Ordinance.pdf](https://github.com/user-attachments/files/18662276/Stafford.County.-.VA.Zoning.Ordinance.pdf)). Observe that Sec. 28-1. - Short title., Sec. 28-2. - Purpose., Sec. 28-6. - Conflict of provisions., Sec. 28-7. - Severability. are not identified as headers, while Sec. 28-4. - Applicability. and Sec. 28-5. - Repeal. are.


### Docling version
Docling version: 2.15
Docling Core version: 2.17.1
Docling IBM Models version: 3.3.0
Docling Parse version: 3.2.0
Python: cpython-312 (3.12.8)
Platform: Windows-11-10.0.22631-SP0
...

### Python version
Python 3.12.8

Start by reproducing the conversion with DocumentConverter() and the attached Stafford County PDF, then inspect how the listed section titles are classified in the PDF-to-Markdown path. Done means Sec. 28-1, 28-2, 28-6, and 28-7 are identified as headers consistently with Sec. 28-4 and 28-5.

Written by the indexing model from the issue text.