I plotted your data for the 3 bulb unit on log log paper and got:
The line is the slope if the data follows the inverse square rule. You can see how much scatter there is, and that the close to the bulb readings diverge from the other data. This isn't much different from the data I got from other sources. And, it is why the charts I made are only good for approximating the PAR - you have to measure in your tank with your lights to get really accurate measurements.
If I were to add your data to my charts I would read off data points from that line, not your actual data. That reduces the scatter a great deal, and makes it possible to do this.