The problem here is that in HTML the <a> tag is not an empty tag. The <a /> usage is not recognized by the browsers.
When the editor received the data, it sends it to the browser and malformed code, like <a />, gets automatically fixed. But in fact, the browser sees <a /> as the opening <a>, with no closing </a>, so the anchor get applied (and split) through the entire document.
Note that the editor will never create such kind of anchor code. It will be always <a name="foo"></a>.
We can hardly fix this issue at our side. You must instead avoid creating malformed HTML by hand as sometimes we're not able to predict the way the browser will behave with it.
