Let's just put this into more concrete terms on some arbitrary efficiency numbers:
Assuming your accessories sap about 20% of your efficiency at idle, a 10% underdrive pulley is going to improve overall idle efficiency by roughly 4%. A 20% underdrive bumps that to 8% at idle.
I think it's fair to assume that the inefficiency of running the belts through a tighter bend is relatively minor - particularly if you upsize one or more of the accessory pulleys being driven.
I think it's also important to note that the improvement I suggest is at idle only. As rpms increase, you get into the accessories like power steering pumps and alternators bypassing/reducing their load on the system. So while one may see a 10 or 20% improvement in efficiency at idle, that won't translate directly to higher rpms - unless something like a clutchless fan is being driven.
Based on that, there should be a measurable improvement, but higher rpm running will limit it.