I would like to know what the consensus is for coordinating the upstream protection device with both the 3-phase and unbalanced damage curve of the downstream transformer? Someone mentioned to me the other day to just coordinate the device with the 3-phase damage curve only. This does not seems okay with me. Any thoughts?

If you assume that you have sufficient unbalanced protection downstream and that failures in the area between the transformer secondary bushings and the downstream secondary protection are sufficiently rare, and you have a magnetically coupled design (no wye-wye or single phase transformers used in three phase arrangement), then the primary side would only see unbalanced faults in two cases. First, if you have an upstream fault that causes unbalanced voltages on the primary side. Usually the analysis for a single phase loss is similar to motors even with fuse protection...if you've fused correctly as the two remaining phases increase to 1.87 times normal current, this is usually enough to trip a fuse within the damage curve. The troublesome case is when you have a high resistance fault though where you do not have complete phase loss...if the fault does not self clear (they usually do), it can confound your protection scheme from operating. Again, careful analysis potentially using something I like to call pen and paper (old fashioned as it may be) is the best way to look at this.

Second, internal faults. In the case of internal faults about the only "saveable" case I've seen is because internal offline tap changer wiring occasionally fails. Usually we've picked this up on oil sampling though if it is saveable. Mind you we have a transformer repair shop within a 100 miles that is very reasonable on pricing to the point where the biggest expense is transporting the transformer. I wouldn't send them anything smaller than 300 kVA though even then. If the internal coils are gone about the only thing you may get is a core that might save you some time getting it rebuilt (the money difference is nominal in the end).

The ANSI C57 Thru Fault curves that you are referring to have a 0.577 shift to the left for line-ground faults on the secondary of a delta-wye connected transformer and 0.866 shift to the left of line-line faults on the secondary of a delta-delta transformer.

The problem is that for each of these unbalanced faults on the secondary side, a secondary winding can experience it's maximum short circuit current and need cleared based on the ANSI curve. On the primary side of the transformer, the primary protective devices are seeing less current on the order of 0.577 or 0.866 depending on the winding. It is best if the time-current curve of your primary protective device can pass to the left of the 'adjusted' ANSI curve as much as possible.

It is usually difficult to pass to the left of the entire adjusted curve so you do the best you can. Having a secondary device's curve pass to the left of the adjusted curve is also helpful. Look at the curve and you will see a change in slope. This occurs at 6X the transformers full load current and is usually considered more of an overload problem than a short circuit problem at that level.