I agree with you that every proprietary software must be presumed to be a trojan with a backdoor but still some critical, low level free software being decades old C codebases with oftentimes millions of LoC has proven to be a double-edged sword where on one hand most of it is super optimized (just compare launch times of lightdm or GDM to e.g. regreet) but on the other hand by pure probabilistics it's more likely to contain some vulnerabilities accumulated over the years of imperfect code reviews.
Sure I believe it's worth hoping and supporting initiatives that might one day bring us to something like RISC-V smartphone with high level of hardware security that'd run something like Alpine (a minimalist distro) but based on Redox OS. Maybe that'll come true in a couple of years.
But right now GrapheneOS even despite proprietary hardware is the best option security-wise, unless you're willing to tinker with hacking together some RISC-V SBC-based device (which might even have better battery life than most smartphones by up to 60%!), but the optimization of basically any software is going to suck so badly. And forget compiling any Rust code on the currently available RISC-V CPUs. want memory safety? pick something with a VM/GC instead.
Apparently there are multiple crates but no official toolchain so unsure how that works in practice. You're still limited to either waiting for hours or cross-compiling though since currently the best available RISC-V CPU is quad-core 2.5 GHz (which still looks hella promising, 2 years ago best we had was 1.5 GHz dual-core). This blog post by Drew DeVault goes into detail of how daily driving RISC-V looked like 2 years ago. I suppose these days it looks noticeably better, especially since Samsung and Apple have been eyeing RISC-V adoption due to ARM consortium doing some monopolistic shit with their licensing. But eventually, so far, not enough critical mass was attained and afaik the whole drama mellowed out a bit.
Regarding the energy efficiency, some experimental units managed to even be manifold better at this:
https://arstechnica.com/gadgets/2020/12/new-risc-v-cpu-claims-recordbreaking-performance-per-watt/
But on the other hand, studies involving some RISC-V models show quite the contrary when it comes to energy efficiency, although the thermal performance is much better:
https://link.springer.com/article/10.1007/s11227-024-05946-9
And below is a screenshot from a comparison by Gary Explains using more microcontroller models. So it really depends on the specific model, but it seems like the design of RISC-V has some solid potential to beat ARM in terms of energy and thermal efficiency.