I would kind of agree in that it currently makes more of an impact in a larger number of games, but the same could be said of other new upscaling techniques that don't necessarily require specialized hardware. And really, the large price hikes for a given level of hardware arguably counteract the performance gains brought on by upscaling, so I'm not sure much performance has actually been gained by the end-user, over what would have been expected without upscaling at this point.
In the long-run, RT may become more relevant, though it will probably be some time before many games actually require hardware RT support to run. In the mean time, RT will create an increased workload for developers as they need to implement and test two different lighting models if they don't want to exclude the majority of their potential customers who either can't run games with RT effects enabled, or can't run them well. Upscaling can only get you so far without looking bad, and the increased input latency and potential for artifacts from frame generation prevents it from being a suitable solution for many games. And by the time RT becomes standard, earlier-generation high-end cards and current-generation mid-range cards might not even be able to run it well. Atomic Heart might have just delayed RT-support at launch because they knew their implementation ran bad on most hardware, and didn't want the performance issues affecting review scores.