The window size is constrained by the point spacing, not by the display configuration. The tool simultaneously loads 9 adjacent bins in a 3x3x3 configuration. Each bin is a rectangular volume defined by the "Neighborhood Width" & "Neighborhood Height" parameters. The tool imposes size limits on the neighborhood to control memory consumption. The neighborhood size is weighed against the average point spacing to estimate the number of points that would get loaded in each bin. To process larger areas, you can thin the LAS data to a higher point spacing using the "Thin LAS" tool, assuming you don't need the 5 cm resolution. Thinning this sort of high resolution data prior to processing will help improve the performance of other downstream processes as well, and it is often the case that there are points which are so densely packed together that they cannot be distinguished from each other.
