Hmm...now I'm getting annoyed. At first glance, I'm actually assuming there is a cube where the yellow arrow points, and also where the green arrow points.
Edit: I think I may have a tiny hypothesis as to why they would present a problem like this. If you look at the green arrow, you HAVE to assume there's a block there or else the structure would be disconnected. Simple.
The yellow arrow, however, is pointing to a spot where they want us to assume there is a cube. But if that spot doesn't contribute towards connectivity or the support of a column, for example, then we cannot assume that it's a cube and not an empty space. Make sense?
Basically, we can only assume a cube is occupying an illusion-like position if it adds towards connectivity or towards support. Thaaaaaaaat's my theory. I could be wrong.
Edit again: But then again, since we have seen evidence of diagonal/corner connectivity between cubes, one may also over-confuse themselves by assuming there is a cube where the yellow arrow points, but not one where the green arrow points...in which case a cube would lie where the yellow arrow points and one could assume that it is connected diagonally to the first cube we see appearing in the back to the left of the column.
Don't read what I just wrote, you may punch a cat afterwards.