There seems to be an assumption by some that the range of a thermal strain cycle must be direct. That is not true.
Even if the pipe "rests" at ambient temperature for a while the total strain range (in this example) is based on the change between 250 to -50. Even if -50 is held for only a short time, the pipe "sees" that strain. (It's like when you dent your car fender, the applied strain is of short duration but the dent remain.)
Yes, CAESAR II only runs from ambient to operating. It's up to the user to build the maximum range (here, either by subtracting the two expansion ranges cases or by running a new case of Operating 1 - Operating 2; of course linear models can be simpler)
And sha (indirectly) brings in the matter of cummulative damage. But even if you hit -50 only a few times, the N you use to set f may increase. You are supposed to convert the N from ambient to 250 (and all other ranges) to an added N for the -50 to 250 case. See eqn. (1d) in B31.3.