Not sure what kind of responses that you are going to get, but my testing loop sounds like it is pretty close to what you have described. When I first tested on this rolling loop, the ride data showed that I was basically either cranking out some impressive watts or damn near coasting. There was very little in between.
I found out over time that there is a cadence and gearing approach that I can use that levels out the power output over this rolling hills type of a loop. I spin faster and grind less over the climbs and then am able to continue pushing coming out of them so that my overall sustained power ends up being higher over the course of the test. In other words, I get less of a power spike grinding up the climbs but then don't have such a low number during the descents and transitions. I find that the only limitation using this method is how fast I can go coming off the climbs and still maintain control of the bike.
On a flat section of road...that I had to drive across the county to find...my norm power and my average power over the course of the entire ride are close to identical. The first time that I tested on this rolling hills loop, there was about a 20% disparity between my norm power and the average from the entire ride. Now that I have gotten better at my spinning and gearing approach, I see that the norm power and average power are getting pretty close, even with the climbs and descents. At this point, I am getting closer and closer to the average power on either testing loop being about the same.
On my CycleOps Fluid 2 trainer, I get a consistent norm versus average power over the same type of test. The average power consistently ranges to be about 7-10% less that what I would see out on the road test though.
I have ended up keeping and tracking both indoor and outdoor test data so that I can gear my training accordingly whether I am riding on the road or inside. When I didn't I found that I was riding way too hard indoors in order to achieve the training power levels suggested by the data from my road tests. After several weeks of that I was seeing some clear signs of overtraining.
Good questions! I will be interested to see what the gurus have to say.