This is a multi-level problem , to be sure. The sequence of events that has to be addressed is:
(1) Be able to specify, when defining a sequence in Creo, the type of "coolant" you want.
(2) Ensure that Creo is outputting you coolant specifications in a unique way to the .ncl file.
(3) Have the post-processor for your particular machine generate the correct M-codes based upon your coolant selection.
Step (1)
For this, you might be able to specify the coolant you want using the two parameters that describe a coolant, the TYPE and the PRESSURE.
Normally, at least for me, the PRESSURE is ignored, it's just the type of coolant that is set and off we go. For what you're dealing with, a combination of the two could make it specific what you are after. I looked at the options for a typical sequence (I'm on Creo 4, by the way) and the lists are:
TYPE -> FLOOD, MIST, OFF, ON, TAP, THRU, THRU_SPNDL, THRU_TOOL, and FROM_RAIL
PRESSURE -> NONE, LOW, MEDIUM, HIGH, AIR, AIR_BLAST, AIR_OFF, RAIL_OFF, NUMERIC
So you probably could use a combo of two of these for each of your unique types, like THRU_SPNDL AIR for your M73, THRU_SPNDL AIR_OFF for M74, etc. Do the same for the other type of air cooling you want, maybe with FROM_RAIL AIR.
Step (2)
If I were trying to do this I'd try some combos, then see what ends up being written to the .ncl file, just to be sure things are getting through, that the syntax is correct, etc.
Step (3)
This is the really tough one. Once you're sure your desired commands are being processed correctly by Creo and generating unique outputs to the .ncl file, you have to now get the post-processor to output the correct M codes to your final .tap file. This particular bit is going to take some doing, and would be hard to explain in a short forum post. It involves programming in FIL code. Folks, particularly PTC, are reticent about doing this stuff for people because the ramifications of mistakes are huge. Crashed machines, bad parts, or both. I've done a good deal of this stuff, but it's always a real chore and debugging is not easy. I've done a lot of programming in general so it's really just another very strange language to me, but for most folks it is probably best to do as suggested previously and contact the AustinNC folks. They are the industry experts, after all.
That being said, if you think you could do some programming, it is very possible to get this to work properly. It'll be a lot of testing, and then having to always remember what the different parameter settings are. I like doing these things myself because then I get to know the machines better and also develop some NC "tricks" along the way.