No it doesn't. The HTML5 parsing algorithm is well-defined, and parsing it according to the spec is not an order of magnitude more difficult than properly handling XML according to its spec. What makes browsers difficult does not come from the part that involves ingesting the initial response payload and building it into a tree in a spec-compliant way—what makes things difficult is all the other stuff involved with creating and maintaining a browser.
"Invalid" HTML5: Please refer to section X.Y for the heuristics to apply in this scenario.
"Invalid" HTML4: Uhh, just do whatever? But probably you need to do whatever the other browsers do.
Implementing the other behaviours is more work than not implementing them, and may be more difficult if your internal data structures don't match those of the browser for which it was originally introduced as expedient or even emergent behaviour.
SGML has no opinions on rendering, since it's just a generic markup language. Nor does it have opinions on "what do with a hr that's a direct descendant of the table despite not being permitted", to pick the first example from the spec (it should be treated as a preceding sibling).