Bon Dia, (as they say in Aruba)
It is important to remember that we are evaluating primary stresses (sustained P + W) differently from our evaluation of secondary stress ranges (thermal expansion/contraction or "displacement"). It is also important to remember how we developed the equation for the allowable stress range limit. It may be useful to look at this post:
http://www.coade.com/ubb/Forum1/HTML/000057.html .
In the B31 Codes the secondary stresses are addressed in terms of the stress range from the coldest temperature expected to the hottest temperature expected. This stress range is due to the CYCLING through thermal excursions (and other causes), and so the Code rules are focused upon fatigue. The "full" stress range combines (ADDS) the thermal excursion from the erected (as opposed to "fabricated", this is North American terminology) temperature to the hottest temperature, with the thermal excursion from the erected temperature to the coldest temperature. So in terms of temperature from your example we will have the "delta T" from the range 85 degrees C minus 21 degrees C (64 degrees C), added to the "delta T" from the range 21 degrees C minus negative 90 degrees C (111 degrees C). So the total "delta T" for the "full" cycle is 175 degrees C. Of course, there will be many "partial cycles" in the life of the piping system and these are addressed by adjustment to the stress range reduction factor, f, as described in B31.3 paragraph 302.3.5 So if you allow Caesar II to calculate the stress range from the "as installed" (i.e., "erected") 21 degrees C to -175 degrees C (if I got my sums right) you will have the calculated stress range for the "full thermal excursion stress range".
So if I understand correctly, your "erected but out-of-service" condition will result in deflections in the piping system that will strain the piping materials "in the opposite direction" from the material strains of the "in-service" condition. The thermal excursion from the "erected but out-of-service" condition to the "in-service" condition will be your "full thermal cycle range".
Now, it is also important (e.g., for hanger design and loadings on pumps and vessels) to evaluate the steady state operating condition (at -90 degrees C) to determine the actual "in-service" or "operating" deflections at that temperature (using the "delta T" of 111 degrees C to in your Caesar II model). Also, for primary stresses (P + W), it is necessary to check the calculated maximum stress against the allowable stress at both the minimum temperature and the maximum temperature - sometimes (not likely in your case) the higher temperature allowable stress will rule.
Pasa Bon Dia (can't wait to get back to Aruba)
Regards, John.