In real life, benchmarks generally only prove how fast benchmarks run.
The problem with all USB speeds as they are given is that they are not the actual streaming speed, unless they've changed that in 3.0 - the speeds e.g. 480Mb/sec is in Burst mode - which means until the device that is sending the data has exhausted its buffer. Once that happens, it slows down. So running a benchmark as you have done proves something, which is that apparently if you keep shoving data at a 2.0 device consistently, it can run in burst mode longer. Having a 256MB file broken into segments from a half k (really? do you do I/O of 512 bytes?) to 8K will probably guarantee keeping the buffer full as long as possible, and as long as you don't do anything else, like scroll a window, or move your mouse, or have google mail running in the background, it will probably keep streaming.
But most computer experts agree on the following things:
1. When using a rotary storage device, the smallest component of any I/O service is the actual transfer time between the device and memory.
2. Typically this is below 15%. Speeding this up doesn't affect the other 85% of the i/O. So if you cut the transfer time in half, you only affect 15% of the overall I/O service time.
3. The best measure of a computer's performance is throughput - i.e., how much work can you do on it.
If I had been running that benchmark, I would have quick-formatted the drives before each run, and run in I/O comparison mode.
