The reason for this is that in order to do well in STEP, there's a fundamental barrier that has to be crossed, between simply being able to solve standard problems using standard techniques, and being able to solve non-standard, abstract, and unstructured problems using a range of techniques. A candidate who can only do the former (in other words, a semi-competent but unremarkable A-Level mathematician) can get a 3, whereas in order to get a 2 or above, he would need to cross the aforementioned barrier, and this is why there's such a gap between the boundaries for a 2 and a 3. Once that fundamental barrier has been crossed, however, going from a 2 up to a 1 simply comes down to gaining experience with problem-solving through practice, and developing the ability to work accurately even in problems which are demanding algebraically or conceptually. This is why the difference between a 2 and a 1 in terms of marks is often not that much.