When I first read this, I must say that I was not totally surprised by its results and conclusions.
This study was done on a group of female students at the University of Cambridge. In the study, they asked practically the same type of question, covering the same material, but in different ways. The students seem to do better when answering the questions when the "... questions are scaffolded...", i.e. it asked the students to answer one piece at a time, leading to the final answer (see the example in the paper. The paper is open access, so you should be able to get a copy of it.).
I find that one of the most common issues when students are given an entire problem in one shot is that they don't know where to start. They have all of these information swirling in their heads, and they don't know which one to use and applicable to answer the question. So having this "scaffolding", where the question asked for something obvious, and then lead the student to another level, certainly might help in guiding the student towards the final answer.
I remember my time as an undergrad at UW-Madison, taking an E&M class with Prof. Bernice Durand, that she had a unique form of assistance during her exams. She actually told us that if we got stuck, or can't answer a question, we could walk up to her during the exam, and asked for hints. Then, depending on the question, she might write something either as a hint, or something to start off. Depending on how much help she gave, the student won't get credit for knowing that part of the solution, but at least, might be able to continue and solve the rest of the problem. She told us that this way, both she and student can diagnose the source of the problem (i.e. say the student just didn't know where to start, but once that is solved, the student was able to carry out the rest of the solution),
I think this is a similar idea to this study. So as someone who benefited from this structure, I can understand how a student might do better when questions are framed like that.