That 2004 file looks a little weird. Some of the conversion might be wrong.
I see in all calibration the max and min multipliers are capped at 1.3 to 0.9. So the multipliers might be added together before beeing applied to main table. On 98 ls1 there are some crank dwell settings that might be not defined on the later bins.

The coolant must be converted to some coil heat soaked scalar matrix, so the pcm predict the temperature of the coils based on coolant, engine runtime and iat. I am sure alot of dwell tables are missing from the files. I will have to start that dwell disassembly to figure all of the unknowns.
The estimated temp of the coils might be used to compensate for the copper resistance change related to temperature. The higher the resistance the more current is transferred to heat , the less efficient the coil gets. The more likely a melt down condition to occur real fast.

The dwell might decrease with higher rpm for more that enough reasons.
I can only speculate. It could be pcm software/hardware limitation of firing more than x coils at a time.
Not enough time to cool the coil between cycles.
Keeping coils colder to extend service intervals.

On a side note truck coils are 2 different designs and are interchangable within the same tune. First design have the mitsubishi star logo and the later are the d585 delco units. They use different lenght spark plug cables due to different height and are only identified by visual inspection. Can`t be figured just by vin number. So the tune might be for one of the coils and not optimized for the other or somewhere in between.