but don't say
anything about width-vs-speed penalties.

Compass generally claims that width itself doesn't cause significant differences in net performance.

I can't remember which issue it was (43 maybe?), but one of the editions of Bicycle Quarterly included testing on the apron of Marymoor Velodrome where the entire range of Compass tire widths performed identically, within the precision of their measurement methodology.

The diff in cruising speed between 40's and 28's looks like 1-1.5 mph.

Fully-rigid bike.

Does this sound reasonable?

I haven't done a rigorous study attempting to only change tire width on the same bike. But when I run Rat Trap Pass ELs (53mm) on my gravel bike, on flat ground, I can't distinguish a cruising performance difference between it and my skinny-tired road bikes (assuming they're also equipped with high-performance tires).
This confuses my riding buddies to no end; there have been group rides when I've ridden my Emonda where people asked me if I was having a bad day, because my pulls weren't hurting them any more than when I previously rode my huge-tired bike.

What I *do* notice having a big effect on paved performance is the style of tire. There was a small period of time last year when I had ThickSlicks (which have a coarser casing and a bunch of puncture protection) rather than Rat Trap Pass ELs on the gravel bike, and it was seemingly around 1mph slower. Knobby MTB tires tend to have an even bigger hit.