I don't think there was an issue with the conversion. If you look at the raw BB-Code it looks more like the article has seen its share of WYSIWYG edits. Since editors don't see the code itself in WYSIWYG editing, you get all sorts of random code fragments, if by chance they highlighted empty space, or removed something that had WYSIWYG formatting.
In any case, this particularly happens when H=1 is surrounded by other block-level code. H=1 is designed to start a new post block on articles, but if the tag is surrounded by another tag that creates a div or any block-level element (like LEFT and CENTER in the aforementioned post), then doing so creates badly nested HTML, which renders (rather kindly, actually) as an alignment issue for the remainder of the article.
I have rewritten a portion of the parser for 3.0.6 and adjusted the headline function so that if it is set to level 1 and detects any surrounding tags, it will neatly close those tags before itself, and reopen them after itself.