Are they all the same soils? I mean if you have a thousand feet of subgrade and it'll be sandy lean clay, and you have another thousand feet and it's silty sand with gravel, you should look at these two regions differently.

If they are all from the same soil type, I'd typically take two-thirds of the average.

f-d

¡papá gordo ain’t no madre flaca!

What is the length of the stretch over which the Cbrs were tested? I would take the Tenth percentile i.e the value which 90% of the Cbr values exceed.

The other option would be to plot a CUSUM for the cbrs and then delineate uniform sections with each section having a different design Cbr.

I would want to increase the CBR values with additional stabilization or treatment before I worried too much about the design.

Personally, I do not like to fill potholes so I design towards to worst case. If the worst case areas can be isolated and improved/stabilized then that will save someone money vs a thicker section. If the owner wants to accept the maintenance issues related to a lower CBR then that is thier decision. The method selection of the design CBR value from a data set is often dicated by the agency guidelines . This can be strict in some areas since the DOTs don't want to be running reparing shoddy work.

For practical reason your design should end up with one pavement design. It's usually not practical to change pavement details in the usual project. Averaging would usually not be feasible, unless a knowledge person would direct the undercut and other fixes as subgrade is exposed.