Seems like the problem is not with the database itself, but with the published results.
I would consults several of the latest reviews to see how the data were treated/averaged/aggregated and then I would choose the one that I like the most.
Looks like at least some of the temperature ranges data are reasonably consistent.