The following test data resulted from adding about 1.5 pf compensation to the output stripline. I did this with 12mm of .085 coax ( center conductor to the output trace, shield to the ground foil, open on the other end). You can use a chip cap if you prefer, but the small coax will do better with the higher RF currents.

This compensation brings up the output across the entire band, and improves the overall efficiency of the module to about 30%, comparable to the silicon modules.

The current PC board rev uses a microstripline stub for the compensation.