Diabetes is a dynamic disease. Although we can use various static formulas to guide insulin dosing such as insulin to carb ratio, insulin sensitivity, and insulin duration of action, many of us also observe that these metrics are not rock solid. They each, in turn, move within their own range.
They can vary from day to day or even by time of day. They may change due to exercise intensity and duration and for many women by their current position in their monthly hormonal cycle. Simply counting nutrition carefully and dosing accordingly does not insure an optimal insulin dose each and every time.
So we are faced with a system with inherent elasticity in the major factors that determine the best dose size and timing. Bernstein has written about his law of small numbers. It basically says that fewer carbs drives taking less insulin which in turn means smaller mistakes.
Making a 10% error on a calculated insulin dose of five units means missing your target by +/- 0.5 units of insulin. A 10% error on a dose of 20 units equates to a possible dose error of +/- 2.0 units of insulin. You can easily see that as the possible dose error, as measured in units of insulin, goes up, so does the risk.
Now, it's possible that a person with diabetes can get very good at dosing for high carb meals. But, with all the moving parts in glucose metabolism, it's highly likely, due to the sheer number of insulin dosing decisions, they will, over time, face a significant under- or over-dose of insulin.
That is a high stakes game I played for many years. Due to a robust habit of fingersticking up to 12-20 times per day before the days of the CGM, I was pretty good a catching errant insulin doses -- until I didn't. Unfortunately the consequences of overdosing insulin can be severe. I am lucky to be alive considering the two or three dicey situations I found myself during my 33 years of living with T1D.
Once I discovered the beauty of a carb-limited way of eating together with its inherently safer nature, I will never go back to playing that high-stakes game. Getting the higher insulin dose calculation right 999 times out of 1000 is not good enough in my book because it only takes one unlucky instance to expose the real risk since the consequence can be terminal.