Thanks, the tiny wheels did it.
I did build real "robot" now with two motors and two wheels, and it showed 9.89m/s or 35.6km/h !!
The whole "robot" two wheels carry the complete weight of 73g, the nail just forces circular movement.
Copying in youtube video text with the details:
It turned out that in previous run only one motor did run. The 2nd was dead due to too hot soldering.
Here both motors (bought a new one) did run, and achieved nearly 10m/s or 36km/h !
With new motor, cabling was a problem. This is first run where centrifugal force did not cut any cables.
Two motors are much louder than one 
The whole "robot" two wheels carry the complete weight of 73g, the nail just forces circular movement.
Now the rowvid.com analysis (20.27s - 21.63s):
9.75 rounds in 1.36s gives 7.17rps or 430rpm of wooden stick.
Radius of 22cm of wheels wrt nail gives 1.38m per round, ppeed now is 9.89m/s or 35.61km/h (!).
Wheel has diameter of 2cm, so with 7.17 rounds of 1.38m this is 157.4rps or 9447rpm at motor shaft (pretty close to data sheet max of 17500 under load).
Last comment on Arduino microcontroller, program did increase PMW from 20 to 255 in 12 seconds, keeps at 255 for 10 seconds and then goes back to 20 in 10 seconds before stopping and LED blinking. In order to use this simple program that has been used in so many testruns now, I did duplicate the PWM signal and both motors are controlled by the same L293D motor output. The second L293D motor driver output is not cabled/used yet.
Hermann.