A quick note because I found the manuscript I was thinking of - Theoretical optimization of multi-echo fMRI data acquisition
If my quick read is right, then the target is ~1.5 x T2* of the tissue, so ~60 is perhaps even on the low end. In other words, the more echoes, the better, and the longer the better (up to a point).
I think the analysis you’ve done there in instructive for this - that last echo is adding a lot of useful information about the BOLD contrast, and therefor helping the fit. I’ll try and take a closer look shortly.