In the early seventies, companies like HP and Wang sold 'programmable desktop calculators' that were really personal computers
Not really, as they stood firmly on the calculator side. If at all, systems like the Cogar 4 and Datapoint 2200 are the origin of desktop computing. Complete units with a CPU, mass storage, CRT display and a full figured typewriter keyboard, all in one unit to be placed on a desk.
The de facto – and indeed de jure – standard interface for computers controlling random equipment was RS-232.
Not really. For one, RS-232 is still a somewhat complicated interface to build (that's the time before integrated circuits) (*1), but equally important, it lacked standardized protocols and application. If at all, HP's parallel HP-IB (HP-Interface Bus), available since the mid 1960s, ruled the area. It became soon known as GPIB (General Purpose Interface Bus), a term coined by companies building compatible devices but trying to avoid the HP name like hell. Later standardization in 1975, as IEEE-488, eased the burden.
Sustainable definition of the HP-IB enabled to not only ease the development of applications, but it also allowed to operate multiple devices on one interface (*2). An HP-IB enabled computer did only need to have a single interface to control (almost) as many devices as needed, thus enabling to handle a whole setup, not just a single instrument. With a predefined protocol device manufacturers needed only to add application specific data/protocol.
Serial in contrast is a point-to-point interface without any protocol support at all. It needs a separate interface for each and every device to be connected, making it quite expensive on the hardware side - not to mention, that there's usually only a finite and rather small number of serial interfaces that can be added to a given computer. On the protocol side applications had to do everything from arbitration and framing to message sequencing and command separation.
Did any of the early programmable calculators have RS-232 ports? (Or if not, why not?)
If they were intended to control measuring devices, a GPIB interface was the way to go. Serial was only good for data transmission. So for example many radio applications used it - and of course terminals. Both not necessarily applications for calculators.
In fact, HP (and many others) even offered serial controllers to be operated via GPIB. Making it easy to extend any device with a GPIB interface to handle many serial connections.
Serial only became a cheap and popular solution when integrated controllers became available (*3). Even then most of the downsides persisted. Still today it's confined to the topic of cheap low level point-to-point connections. The stuff hobbyists prefer as they can replace functionality by investing their time to add hardware hacks and software layers.
On professional measurement equipment GPIB is still today (2019) the most important interface, even though many affords are made to replace the hardware layer by an IP or USB based connection.
No company in the measurement business could afford to deliver devices without GPIB from the 1970s until today. Just do a search for PCIe based GPIB controllers and you'll be in for a surprise. That market is so big, that many manufacturers crank out new interface cards whenever there new desktop systems become available. Similar on the device side.
Of course, all of this happens more on the professional side, where devices easily carry 5 digit price labels, but not so much in the sub-1000 USD realm where hobbyists fight for pennies.
HP reused the GPIB idea even as the serial implementation HP-IL, for their pocket calculators, when they became powerful enough to act as controllers for other devices. Most notably the HP-41 series.
*1 - For a serial interface bit timing circuitry, sequencing logic and a shift register is needed for sending and the same again plus additional word synchronizing on the receiving side. All plus the optional parity logic. Quite a lot. A parallel interface in contrast is just a set of two latches (one can even be just a buffer) with clocked input and output enable. Quite less and much more simple hardware.
*2 - Features that enabled the success of USB half a century as well.
*3 - Serial interfaces of the 1960s were shoe box sized or larger.