Just goes to show you how different we all are. I used to run into problems with Regular and Lente because they would never produce the same activity and duration curves twice in a row.
Sometimes Lente would last me 26 or more hours and sometimes less than 12.
And variation in peaks were a big problem. Sometimes the Regular and Lente would unexpectedly bump into each other when the graph said it wouldn’t happen.
Fun times, fun times.
I could believe there might be a tiny tail on NPH that lasts that long, but most of it is long gone by then. I think the same is true for that graph re: R, at least for me, in that the tail seems overemphasized—most of it is used in the first 4-5 hours, with a strong peak at around 2 hours, and what’s left after 4-5 is marginal. Maybe that’s because I wasn’t using it with a 24 hr basal though (was using R and NPH or R and Lente back then), so those tails were in fact filling in basal gaps and would be much more noticeable if added on top of existing basal.
That makes sense.
This might work pretty well for what I am seeing.
as if we don’t have enough alarms!
I am arriving after the battle 
For all it’s worth, we use an ICR. We typically use our regular ICR for all meals that are mostly carbs, up to 150 carbs (the most we ever had for carb-mostly meals). For other meals that are high in carbs but also have a lot of proteins and fat, we use 100% ICR upfront, then when on MDI, a second injection of 30-50% 1.5 hours later, and on the pump, a 2-hour extended bolus of 30-50%.