As another commenter pointed out, even if the gate isn't ready yet, there's always room to wait somewhere on the taxiway (this has happened to me a few times recently). They'd rather wait on the ground with the engines idling than circling around in the air with the engines running at speed. Plus, it's not like gate availability is affected by schedule padding anyway; there's only so many gates at an airport, and they always are trying to turn over planes as fast as possible.
It's like if someone said, "going a little slower on the highway [typically ~60 mph+] saves on fuel" and someone pedantically ButAkshully'd that cars are more efficient at 40 mph than 30 mph because of engine performance.
Edit: Ironically, I'm the one normally promoting my "Scylla-Charybdis heuristic", that you should be able to identify when tweaking a parameter goes too far. But here I think the context of the tweak was clear enough that this caveat was a distraction from the core point.
In practice they save fuel by flying slightly slower. But, their are many other considerations like weather that cause airlines to fly slightly outside of optimal altitudes and speeds.
It's counter-intuitive, but an aircraft can be more efficient at higher airspeeds. The reason for this is that, unlike a ground vehicle, an aircraft in level flight must expend energy generating lift to balance its weight. This loss of energy shows up as lift-induced drag [1]. As airspeed increases, the losses to lift-induced drag decrease, while the losses to aerodynamic drag increase. Where the point of minimum total drag lies depends on several factors including weight and altitude. See [2], which contains a chart for a generic aircraft which is 5% less efficient at Mach 0.72 than it is at Mach 0.80.
1. https://en.wikipedia.org/wiki/Lift-induced_drag
2. http://atmseminar.org/seminarContent/seminar11/papers/481-Je...