have bandwidth that is some % of carrier frequency,

In my limited ham radio experience, I’ve not seen any antennas nor amplifiers which specify their bandwidth as a percentage of “carrier frequency”, and I think that term wouldn’t make any sense for antennas and (analog) amplifiers, since the carrier is a property of the modulation; an antenna doesn’t care about modulation, which is why “HDTV antennas” circa 2000s in the USA were merely a marketing term.

The only antennas and amplifiers I’ve seen have given their bandwidth as fixed ranges, often accompanied with a plot of the varying gain/output across that range.

going up in frequency makes bandwidth bigger

Yes, but also no. If a 200 kHz FM commercial radio station’s signal were shifted from its customary 88-108 MHz band up to the Terahertz range of the electromagnetic spectrum (where infrared and visible light are), the bandwidth would still remain 200 kHz. Indeed, this shifting is actually done, albeit for cable television, where those signals are modulated onto fibre optic cables.

What is definitely true is that way up in the electromagnetic spectrum, there is simply more Hertz to utilize. If we include all radio/microwave bands, that would be the approximate frequencies from 30 kHz to 300 GHz. So basically 300 GHz of bandwidth. But for C band fibre optic cable, their usable band is from 1530-1565 nm, which would translate to 191-195 THz, with 4 THz of bandwidth. That’s over eight times larger! So much room for activities!

For less industrial use-cases, we can look to 60 GHz technology, which is used for so-called “Wireless HDMI” devices, because the 7 GHz bandwidth of the 60 GHz band enables huge data rates.

To actually compare the modulation of different technologies irrespective of their radio band, we often look to special efficiency, which is how much data (bits/sec) can be sent over a given bandwidth (in Hz). Higher bits/sec/Hz means more efficient use of the radio waves, up to the Shannon-Hartley theoretical limits.

getting higher % of bandwidth requires more sophisticated, more expensive, heavier designs

Again, yes but also no. If a receiver need only receive a narrow band, then the most straightforward design is to shift the operating frequency down to something more manageable. This is the basis of superheterodyne FM radio receivers, from the era when a few MHz were considered to be very fast waves.

We can and do have examples of this design for higher microwave frequency operation, such as shifting broadcast satellite signals down to normal television bands, suitable for reusing conventional TV coax, which can only carry signals in the 0-2 GHz band at best.

The real challenge is when a massive chunk of bandwidth is of interest, then careful analog design is required. Well, maybe only for precision work. Software defined radio (SDR) is one realm that needs the analog firehose, since “tuning” into a specific band or transmission is done later in software. A cheap RTL-SDR can view a 2.4 MHz slice of bandwidth, which is suitable for plenty of things except broadcast TV, which needs 5-6 MHz.

LoRa is much slower, caused by narrowed bandwidth but also because it’s more noise-resistant

I feel like this states the cause-and-effect in the wrong order. The designers of LoRa knew they wanted a narrow-band, low-symbol rate air interface, in order to be long range, and thus were prepared to trade away a faster throughput to achieve that objective. I won’t say that slowness is a “feature” of LoRa, but given the same objectives and the limitations that this universe imposes, no one has produced a competitor with blisteringly fast data rate. So slowness is simply expected under these circumstances; it’s not a “bug” that can be fixed.

In the final edit of my original comment, I added this:

Radio engineering, like all other disciplines of engineering, centers upon balancing competing requirements and limitations in elegant ways. Radio range is the product of intensely optimizing all factors for the desired objective.