That's just not realistic, I don't think. Series S has pretty much the best CPU in the console space. It's actually overstocked with CPU grunt relative to its other components, and it has relatively fast RAM, a single "cluster" and shares a silicon die with its GPU. It also has hyperthreading.
A78C is impressive, sure, and at the same frequency it could compete or outperform, but power limits put definitive physical caps on speed. The CPU basically has to be the same speed in TV and handheld mode to not mess with game logic. So the speed cap put in place by the CPU's access to power while on battery is in place for all scenarios. That's probably not going to be high enough to scrape Series S. A significant portion, well over fifty percent of the way there, I'd say so, but I wouldn't call it very similar in performance.
What it has though is more RAM, Tensor Cores and the FDE. Most decompression should be going through the FDE, video decode and encode should go through their dedicated blocks, that's a little weight taken off the CPU. Tensor cores can perform some typically CPU driven tasks, in theory, and more RAM means you could, for instance, make up for the CPU speed in some circumstances by storing precalculated figures in a table in RAM and calling them rather than having the CPU do everything. Not similar, but there's tools at hand if a game is struggling.