In adjusting the map values it seems to me there should be some pattern to the cell values. With this in mind I sucked the InjDur table from the stock map into Excel and did a simple calc.
I expected to find some sort of linear or exponential increment between adjacent cells but to my surprise I found the numbers all over the map (pardon the pun)
For example across throttle opening Row 7 I get a difference of;
3.4%, 1.7, 0.9, 3.0, 2.4, 1.8, 5.6, 7.1, 13.2, 12.8, 0.0, 1.7,1.4,9.4 Percent
Down the 2000  rev colunm I get;
0.4, 1.0, 1.7, 0.5, 15.5, 18.7, 2.8, 6.4, 15.6, 25.2, 16.7, 24.6, 38.0, 72.4 (oops that's more than 100% but you can see the relationship)
It's more than my tired old brain can handle :(

There is no formula. You just can't calculate a combustion engine, it's far too many parameters. If this was not the case, the ECU could theoretically be a bunch of passive components.

For example, my exact setup has a positive harmonic (likely due to lengths of exhausts, speed of sound and volume of air filter box) at 4500 rpm - making it breathe better there than at 5000. So the InjDur at 4500 rpm is higher than at 5000 for the same target lambda.

What you can do is make sure no cell stick out too much. The graph should look like you put a cloth over the map. I think that was Hubert's words, maybe a bit distorted.

Raz, I accept your explanation, I can see that your map would not work for me. I can accept that the a highly tuned engine breaths differently across the rev range but I would still expect some sort of pattern if not between adjacent rows at least between adjacent columns.
The pattern I see looks like I have already messed with it  :( 

Regards
Roy

I guess you are right, maybe I over-reacted against the word "formula". I think what you are after is best viewed in a 3D graph of the InjDur table. A good map is probably pretty smooth with no isolated spikes. I would have the mentioned spike at 4500 rpm but it should be repeated at lower throttles, not just WOT.

Raz,
       Ok, sorry about the formula thing, it was asking a bit much. I think I will try smoothing it out a bit, knock off any peak and fill in the valleys, there's just a few spots that look a bit out and give that a go. Once I get it working better I will try out my majic TPS offset box so I can shift up or down a row or two on the fly (change the injection time without changing the throttle), I guess that's what you do with the optimizer in open loop.
Cheers  :)
Roy

Update:
I stole a table from Raz and copied it into my map.
Without even testing it I took off an a short trip.
The bike purred along like a contented cat.

I'm quite happy with my RazMap  :D