I think III would be better titled "Post-Hoc Hypothesizing," i.e. where you create (or change) your experiment's hypothesis after the experiment - clearly that is wrong. "Post-hoc storytelling" sounds like what you do in the "implications" or "next steps" part of a paper. Also, exploratory studies have different standards, right? (I've never done an exploratory study.)

In the graphic, I thought "Non-Publication" and "Partial Publication" referred to experiments, i.e. where you run 4 experiments and only report on the last one, which "worked" -- clearly that is wrong. However they are talking about data, which is a bit trickier. In some cases you can't just upload all your data onto github; it may take a while because some sponsor or corporate office needs to review it before it can be released. A researcher may refuse to release their data until they finish writing a set of papers based on it. There may also be distribution limitations, or licensing requirements related to the data. I'm not trying to defend withholding data, just that these are the problems that need to be addressed.

Finally, to extend the metaphor, the "publish or perish" mentality is this image's Satan.