Thanks Lights. The odd thing here is the very low levels on your Y scale.
Despite this your ETC shows a very well achieve 70dB between peak and background noise. You could clear up the fan issue easily. Use the Sig Gen. Tick the Frequency follows Cursor box. Sine Wave. Slide the cursor over the LF ridges, the Sine will generate that exact frequency. If there is a resonance triggering the fan rattle you will easily find it.
The non linear slope, particularly on the newly introduced ARC resonance will give false ideas about it's length. The original calculation was over the very flat first 17dB or so. The slope accelerates greatly after that. REW itself suggest about 525mS. I am sure you would have noticed 900!
I believe Jens central point was to be careful in extracting Decay information casually from graphs. Or beware of the unseen. A point well made and well accepted I hope. In the past I simply assumed that EDT, often the first stat presented, was the best choice. It so isn't.