What version of System Info is that? If it's the 68k version (3.1 I think?) it may be worth running these tests with PPC-native software, just in case the 68k version is leaving some performance on the table.
That said: This is genuinely a great showing for SD card adapters, and the mSATA adapter is arguably nothing to sneeze at compared to the stock disks.
w/re 128GB: it might still be worth doing, and this line of investigation is probably still worth having for systems that do support >137GB. The TiBook/1000 for example addresses a 160-gig disk I have fine and a 256-512GB m.2 SSD would be a hilarious, if a little overpowered for it. For systems that don't support LBA48 you can just run your disk as the first hwoever many gigs are recognized and leave the rest of the room for wear leveling. Probably worth under-provisioning a bit for, say, OS X systems since OS X didn't support TRIM until well into the Intel era.
A thing I'd be interested in seeing, if it's posible to test, is how any of these solutions do under sustained workload. Something like capturing a DV tape off of firewire would be an interesting stress test. (Granted, you can just use a desktop and a PCI SATA card and a regular SSD for that, so whether or not that's "important" to test is... an exercise left to the people who have all the parts, which I don't right now.
That's a great point. If an m.2 SSD is faster than an mSATA one it's just because SSD tech got better. If I remember correctly, SSDs reaching that peak 6 gigabit number happened a bit after mSATA fell out of style.
Agreed completely, to be honest, I think the only way anyone will "experience" the numbers AlpineRaven's benchmarks show is if they either use OS X or if they... run a benchmark.