Guessing from the time to load scenes 2-4 from SATA SSD they don't read more than 3.5 GB; that should be about or under 1 second for PCIe 3.0+ SSD. Seems like 2 to 3 seconds needed to load from PCIe 3.0/4.0/5.0 SSD are mostly spent on overhead, with differences in read speed hidden by overlapped I/O.
If you think about it, when you need to completely reload, say, 8 GB of data in VRAM and a few GB in system RAM, you should be able to do that in 3-4 seconds from PCIe 3.0 SSD; and probably 2-3 with DirectStorage compression. Everything else is just lack of optimization or overhead, filesystem (lots of small accesses instead of a few big sequential reads), inefficient/slow decompression, not doing enough things in parallel, overhead (e.g. drivers), plus time of game state reset.
Setting aside other issues with Forspoken, it looks like DirectStorage fell victim to... optimization and great performance.
I don't play a lot of games, but from my PCIe 3.0 SSD, in a few other AAA releases loading a save takes 10-20x the time Forspoken needs for the same (even when changing environment).
So I guess what happened is MS wanted to show off with DirectStorage, made the developer optimize, optimize and optimize some more. They shaved time from, say, 20 seconds to 2, then applied DirectStorage compression to reduce it to 1 second, made it smaller than CPU/GPU loading overhead... and it turned out that, on fast disks,
