I don't agree. The wiki states that "Shown in the table are the measured values with screen switched off and all possible optimizations" and yet I've proven that, for the specific benchmark, you can get identical behavior, but improved performance, by actually enabling "all possible optimizations" on a SCPU.
Yes, it's easy to intentionally modify the benchmark to make it blow up with the BASIC optimization setting on a SCPU, but I fail to see how that relevant.
The Chameleon may be easier to use, much like a 68k Amiga accelerator is easier to use than a PPC one, but I wouldn't call my 68060@80 Apollo superior compared to my 603e@200 Blizzard since the PPC can easily run circles around the 68k by using appropriate software.
On the other hand, any C64 accelerator is rather pointless since 99% of the C64 software relies on specific timing and/or a fixed frame rate and doesn't benefit from a faster CPU.