It looks to me like you are leaving too much of the task to Galapagos. I would generate the five squares in random sizes to meet your criteria *(“the sum of their area should not exceed 75 percent of the area of the site”)*, then **use Galapagos to do nothing but** position them randomly within the site boundary and without overlapping.

I see no reason for using `Rotate Plane`, especially with a choice of either 0 or 1 radians?

I see no reason for using `Scale` since random sizes can be done when creating the squares *(rectangles)*.

.

**P.S.** I started puttering without a plan and got to this, which is **not a very good algorithm**.

Galapagos_2021Nov26a.gh (18.1 KB)

Seems to me now that an iterative approach that places one square at a time might work best, with the squares sorted by size to place largest first, smallest last.

Similar to this, which places rods at random angles without colliding:

**P.P.S.** This algorithm is still flawed but the yellow group aligns each rectangle to it’s nearest edge, which might be useful.

Galapagos_2021Nov26b.gh (22.3 KB)