Here is a tutorial I wrote a while back on dialing in base maps: http://msioc.com/forums/viewtopic.php?f=23&t=266
There is a section on maf scaling in there that might help a little. However, it is a tad outdated as I have learned quite a bit more about our ecu's since then. I usually start by getting the CL portion dialed in really well with the maf tab in romraider. During every iteration, check the graph of your maf table in ECU flash to make sure it is a smooth curve. Sometimes its hard to hit certain spots in the MAF table. If you get certain spots dialed in really well, but theres a bump in the curve caused by a spot that has not been adjusted due to lack of data, you can usually just use some interpolation in ECU flash. Another way to do this, as well as smooth out your curve, is to graph your maf curve in excel, and then add a polynomial trend line, setting it to 5 or 6. Check the box that says "display equation on chart". You can enter this equation into wolframalpha.com and have it solve for the g/s value at the voltage point that lacks data. I am sure theres an easier way to do this in excel, but I'm no excel genius.
Once I have the CL portion complete, I then use a combination of the OL spreadsheet, and manual analysis inside airboys spreadsheet's afr interpolate tab. Fast polling has made this process WAY easier than it used to be. Another thing I do is graph logged WBO2 data compared to final fueling base, as NSFW mentioned. This is still a pretty gray area in subaru ECU's, but it will give you an idea of whether your maf scaling is off in a spot, or if your ecu is just acting on your target AFR's.
Once I get the OL portion within +/- 2% error in WOT, I move on to the engine load comp table by using this spreadsheet: http://www.romraider.com/forum/viewt...hp?f=32&t=5483
If I notice a correlation in AFR error outside +/-2%, and varying IAT's, I then touch everything up with the maf IAT comp table.