You should never lug the engine. This puts extra stress upon it. The car guys said you should drive at low engine speed in the highest possible gear. Shifting up to a higher gear resulting in lugging the engine is not the highest possible gear. Perhaps “possible” was not the right word. 
Perhaps I shouldn’t have used lug (then again, 35-50MPH in fifth gear seems pretty low to me.) My point is that there is a point of maximum efficiency and engine efficiency is not linear and is by no means inversely proportional to engine speed.
While I don’t have a vacuum gauge on my truck, I’ve heard of people successfully using a vacuum gauge to maintain gas mileage, and it makes sense that, given a constant MPH, BSFC would be highest when manifold vacuum was greatest (or manifold pressure was lowest, depending on how you want to look at it.)
Wouldn’t pumping losses also be much higher at higher RPM?
OK I went out and did some experiments today.
Conditions:
Air temp 82F road: level, and steady hill. car setup: A/C off, cruise control set, down shifting auto trans to lower gears to observe instant gas mileage readings
Run one:
40MPH up a grade
4th gear = 21.2MPG
3rd gear = 19.8MPG
2nd gear = 16.5MPG
Run two:
70MPH on flat freeway
4th gear = 25.3MPG
3rd gear = 21.8MPG
It seems pretty obvious to me, higer gear = better fuel mileage.
It was obvious without your numbers too. These results aren’t terribly instructive, I’m afraid. What we need is to compare different mileage readings under differing load conditions while holding RPM constant, e.g., up a hill at 40 MPH in third vs. level run at 40 MPH in third; that sort of thing. What you’ve done is only to compare differing engine RPMs under identical load conditions, which doesn’t tell us much.
Nevertheless, kudos for taking the experimental approach.
QED I’m not going to run out and do two 40MPH runs, one on a hill on one of flat ground to determine which getst the best mileage, since I already know the answer, you get better gas mileage on flat ground. It takes less energy to drive on flat ground then it does to go the same speed up a hill. I mean what would be the purpose of this experiment? 
If you are still not sure of the answer, ask anyone who rides a bicycle if it takes more energy to climb a hill or ride on flat ground.
What I did was try to answer the OP’s question
Silly me I thought that I did that.
Yes, you did, sort of. Thinking about it, though, I think the OP really wanted to know where in the range of RPMs for a given gear does the best fuel economy occur. My suggestion didn’t really address that, either (and yes, I did know that running uphill uses more energy than level running ;)). I suppose graphing the fuel economy for several data points for each gear under identical conditions would answer that.