Vehicle fuel economy is notoriously difficult to measure. Even the Big Guys (car companies/EPA) have a difficult time doing it. The real world has so many variables that official, legally-binding fuel economy tests aren’t even done out there; they’re done in a lab, with the vehicle on a dyno. Everything about the test is controlled, in order to make the test as repeatable as possible:
-the vehicle is held at a controlled room temperature (something like 24 hours) before the test to ensure that its starting temperature is the same every time.
-the tires are filled to a specified pressure with a NIST-traceable tire pressure gauge.
-the vehicle is placed on a chassis dynamometer (wheels on rollers with computer-controlled drag torque) using a vehicle mover (the car isn’t started at all yet). The combined aero/wheel drag characteristics for the vehicle are entered into the dyno’s computer. Those drag characteristics are measured with a carefully-controlled, real-world coast down test. This is done on a long, flat stretch of road (e.g. an airport runway); the coast-down test must be run in both directions to eliminate any effects from slope, and there’s an anemometer on site (or in the best case, on the vehicle) that accounts for local wind variations. Incidentally, this coast-down test is where Hyundai screwed up a few years ago, resulting in overly optimistic MPG ratings on new-car window stickers (and a whole bunch of owner compensation).
-The driver starts the car and puts it through a very specific driving cycle (either highway cycle or city cycle). He watches a computer screen next to his window, which shows his vehicle speed along with error bands; he has to keep his speed inside those error bands for the duration of the test, or else the test gets voided. The bands are pretty narrow, and it takes some experience to be able to complete a whole test successfully.
-The dashboard fuel gauge doesn’t matter. In fact, they don’t even measure fuel quantity directly; they measure the carbon output from the tailpipe over the duration of the test, and infer fuel mass from that.
So, back to the OP’s van. There are so many variables out in the real world (including a fuel gauge with coarse gradations and unknown linearity/repeatability/precision) that it’s impossible to back up any claim with a single round-trip. It’s entirely possible that the aero drag characteristics changed because of the differing ride height; it’s also possible that there was a slight wind in effect (5MPH steady wind means your airspeed is 70 going one direction, 60 in the other), and it’s also possible that the OP drove the vehicle differently on each leg of the trip. The tires may have been cold/low pressure when he began the first leg of the trip, and then much more warmed up (higher pressure) for the return leg.
Having said all that, here’s one other interesting thought:
Gasoline engines are substantially more efficient at higher loads. For a lightly-loaded van, it takes X1 engine power to get up a given hill, and it takes Y1 engine power to overcome aero drag when going down that same hill (Y1 < X1).
Now do that same hill again with an extra 3000 pounds. What happens?
-The engine is making X2 power on the uphill climb (X2 > X1), but it’s also more efficient. When going back down the hill, the extra 3000 pounds means the driver is using less gas (Y2 < Y1). The net result may in fact be less total fuel used: the total energy pissed away to aero/tire drag is the same in each case, but in the latter case, the net efficiency of the engine may be better.
This squares with my understanding of why hybrid vehicles like the Toyota Prius can be so efficient. They have regenerative braking, which helps, but the energy recovery from that is actually pretty small. An additional factor is the ability to use the battery to trim the engine output, allowing it to operate in efficient (high-load) modes even when the wheels don’t need all that power. If you’re into electronics, then you can envision this arrangement as being a bit like a switched-mode power supply, in which the Prius battery is analogous to the SMPS’ capacitor: the engine can operate in a high-efficiency/high-power mode or a low-loss/low-power mode, and the cap smooths out the output. Something similar could well be happening to the OP’s van on those hills, where the engine’s excess power output gets stored as gravitational potential energy and then released on the downslope, allowing his van’s engine to operate in a similar way.
So I think it’s theoretically possible that the OP is enjoying better fuel economy with a loaded van on his hilly drive, but I wouldn’t trust the dashboard fuel gauge to make that determination. Moreover, even if it’s true, it would only be applicable to slighly hilly highway drives. Mountain passes with steep downhills will require braking, which would ruin fuel economy. And city driving likewise requires lots of braking and acceleration, and that’s where the extra mass will absolutely kill your fuel economy.