There are JEDEC speeds, the official standard. With them, pretty much any lower speed than the rated speed is supported, pretty much by design.
XMP or other extended profile settings are separate and explicitly put into the memory module, each of them as a list of settings. This is only done for very few, specific speeds & latency combinations.
And typically, the JEDEC profiles are from what the chips were officially produced at. The final vendor then tests those chips for overclocking. Since they need to do their own testing and overclocking validation anyway, they often use older chips that were sold for lesser speeds, as long as they have the right overclocking potential (hence lesser JEDEC speeds compared to what the highest available JEDEC rating below the OC profile would be).
There can often be gaps in the supported extended profiles where combinations below the 1 communicated rating are not guaranteed. And most vendors do not communicate the JEDEC rating of their OC memory, below which downwards compatibility is usually rock solid.
TL;DR; if you don’t intend to run at the rated overclocking settings in a board that explicitly supported them, ignore anything but JEDEC ratings.
And on modern, mobile laptops you are pretty much not getting XMP or memory overclocking. Modern platforms support various power saving techs, like for example dynamically downclocking the memory and “reducing” latencies when its less needed etc. On desktop mainboards with overclocking, those power saving options are typically hard disabled, because they don’t care about energy efficiency in the first place. And if they are there, they are typically hard broken and will cause all kinds of instability (like my board, which ignores any official specs of the CPU, overclocks it by 2000 MT/s by default, even with JEDEC selected and hangs during memory training or looses all memory integrity to the point you cannot even boot linux anymore, if any of the power saving options standard for a laptop are enabled).
Because the power saving options are tested by the CPU manufacturer against their official, non-overclocking specs and the mainboard vendors have to mess with all kinds of internal settings to make the memory overclocking seemless for the user (overvolt many components, lie to others). So if a platform has memory overclocking, it likely has terrible idle and sleep power consumption and it would be tons more complicated to even just have a platform that can do either. Let alone both at the same time. And Framework is not winning any awards with their idle power consumption numbers to begin with. And that is with power saving tech enabled.
HWInfo for example can enumerate the various different profiles supported.