Interesting idea. A few things though, after trying to plug in the new corrected data 15h42m cycle, I'm getting some strange compounding error discrepancies between my hand calculations and your page.
All times posted are Cavern Time = Mountain standard (GMT-7)
[Old data]
'Sun Mar 4 03:19:00 2007'
Confirmed
'Sun Mar 4 19:01:00 2007'
Confirmed
'Mon Mar 5 10:45:00 2007'
See correction notes
Error correction starts here based on the "10:43" average appearance (ki time) when the portals started to appear. I've included here the new corrected portal appearance time for reference, and their subsequent predictions based on the new day length.
Mon Mar 5 10:43:00 2007'
Confirmed
'Tue Mar 6 02:25:00 2007'
'Tue Mar 6 18:07:00 2007'
'Wed Mar 7 09:49:00 2007'
'Thu Mar 8 01:31:00 2007'
'Thu Mar 8 17:13:00 2007'
'Fri Mar 9 08:55:00 2007'
Since all the times here are Cavern time, I plugged in the new day length, and the start time (I assume its the basis for calculating the next portal time) and I end up with some very odd results.
Also, the problem with predicting so far into the future is that we still don't have a precise enough handle on the exact time between portals. We have it close enough to make a few days worth of predictions, but compounding error can really toy with us if we are not as precise as we can be.
Another thing. I have NO idea how tomorrow's downtime will affect any of these results. Its possible that when the servers go down, all times will be reset... in which case we'll have to start from scratch and find the first portal again.
Edit: Found the problem. Your "midnight" and "portal appearance" values are listed in the wrong columns. The math is good but the placement is not.