I defined the cell as in the size of the full matrixed sub-structure (pattern 1). So the beginning feature is driven by CELL/2 relations for the features.
Tube is the diameter of the tubes, rods, or whatever you want to call them. I found it easiest to manage them as a ratio to the cell size.
Interesting observation in the graphics performance. I am not sure what that would happened unless for some reason the graphics engine is still maintaining the quilt definitions.
Maybe we have peeked the interest of some PTC engineers to review these models and what is going on in the background. We have certainly provided enough models for some serious investigation.