Here's my thought: the less insulin you use to test your ISF in a controlled test, the more that basal discrepancies and other issues over the time period you test, will unduly influence the result either way. Not good.
Then again, you've probably got some idea of your ISF already and whether you are sensitive to insulin or not. The more sensitive you are, the more substantial .5 of a unit might be as a determining factor - for someone really sensitive it might be viable to do it this way.
Here's my example: A unit only drops me about 20 mg/dl, half a unit would only drop me about 10 mg/dl, it would be very hard to pick up that drop over the few hours I'd need to test it. Given tester inacuracies and all the other factors, the smaller numbers involved would make this pretty untenable as a diagnostic tool.
I worked out my ISF through trial and error with real correction needs, but it ended up being very close to what you get with the
1800 Rule and Correction Factor