That example was built to show how deadweight load can be shifted.

I would agree that a positive deadweight load on a restraint would not be optimal. If I have three supports - A, B, C - the sum of their deadweight loads would be equal to the total system deadweight. If A=-500, B=-250 & C=-1000, the total weight of the system is 1750. Now if B was +250, then A+C=-2000. So the remaining supports must also include the extra weight. A or C or both will be a bigger support than in the "optimal" configuration.

Yes, the initial +237 is smaller in magnitude but the -495 has a better looking sign.