Nested lists of roman numerals that differ in casing fail to parse properly

If I try to parse this:

1. **Introduction to Legal Principles**
   1) Historical Context
   2) Core Doctrines
      i. Doctrine of Precedent
      ii. Doctrine of Equity
   3) Sources of Law
2. **Civil Procedure**
   A) Pleadings and Motions
      a. The Complaint
      b. The Answer
      c. Pre-trial Motions
   B) Discovery Process
   C) Trial Proceedings

I get the proper AST:

However, if I change the list before the roman numerals in lowercase to a list with roman numbers in uppercase like this:

1. **Introduction to Legal Principles**
   I) Historical Context
   II) Core Doctrines
      i. Doctrine of Precedent
      ii. Doctrine of Equity
   III) Sources of Law
2. **Civil Procedure**
   A) Pleadings and Motions
      a. The Complaint
      b. The Answer
      c. Pre-trial Motions
   B) Discovery Process
   C) Trial Proceedings

the AST generated is wrong, as it didn't create a third level for the list with lowercase roman numerals:

Reproduce both Markdown snippets and inspect the generated AST, focusing on nested ordered-list parsing and differing roman-numeral casing. Done means the uppercase outer roman-numeral lists still produce a distinct third-level list for lowercase roman numerals, matching the first example’s structure; add or run a regression test if the repository provides one.