In the Legacy XML 1.2 filter, Embedded Content Processing using RegEx rules only affect text nodes and not child elements.
For processing embedded HTML code, in the XML 1.3 filter, it makes sense to handle all child nodes as one chain of text, basically like a CDATA section.
This approach does not make sense if you want to deal with mixed-content elements and handle certain non-XML-compliant placeholders like {variable}, [variable] etc. etc. as inline tags.
Please make sure that the future XML 1.3 filter will enable correct processing separated for text nodes and child element nodes, which is absolutely needed for RegEx Embedded Content Processing.
Otherwise maintaining/supporting the Legacy XML 1.2 filter indefinitely would be another option.