I have done a bit more work on my test rig. I have a much bigger power supply and have fitted a 30v flyback to my (very old) test ECU. This has improved things significantly (test done at 3bar rail pressure and 14.0v injector power supply):
The close time of the injector is now much quicker than the open time as one would expect. I can now use the data to derive some calibration parameters:
The injector size is the gradient of the linear region of the above chart. Converted to the usual units this is 583cc/min.
The "simple" dead time is the value of pulsewidth when flow is zero. In this case, this is 307.4us.
The injectors won't operate linearly below about 2ms. Therefore if the ECU is calling for pulsewidths less than this (for example at idle) the results will be extremely poor. The simple calibration scheme can't get around this.
Obviously more tests are needed for other supply voltages. If the simple model is accurate, the gradient does not change with voltage. I am still working on deriving values for the "standard" injector model.