The grid actually works on a principle called "pull to left." It was a UX decision in the end. "Left" is because the grid is designed horizontally (but should be made an option to use the other axis very easily), and "pull" is because we found (after much usability testing) that if you don't have any force in a drag and drop system, items will involuntarily continue to move away from each other as you move them around and collisions occur.
I get where you come from with the jQuery plugin. However, due to many reasons (including time and reducing in-house complexity) jQuery was advantageous for its build-in drag and drop capabilities. But the good part is that we still separated most of our logic from the jQuery plugin, into abstract positioning. The jQuery plugin just transforms grid data into pixels and lays them around. There are many reasons why this is cool. 1. Very fast unit tests ran with Node, 2. Grid positions can be generated server side, etc.