Citations to data sources where you got the numbers from would be polite to those authors..
Then I replaced ranges with midpoint values.
This is OK, and the best you can do in Excel in any case.
The linear lines that I fitted are an alternative to a single logarithmic fit …
I agree a logarithmic fit is not appropriate unless you have data from a single population, which this clearly isn’t. A linear fit isn’t necessarily better, more on that …
The fit below excludes fossils discovered in this century. When you include the hobbit and Naledi, the R-squared statistic drops to 0.5
Definitely not the same population. It would be good to include those points on the plot, and you can leave them out of the regression lines.
Here is the real issue; your choice of what data to include and where to split the data for two regression lines is crucial. It would be good to should all the data you have first, then make a second plot leaving some data out. This would be more transparent to what you are actually doing and the choices you are making. There must be multiple measurements available for some species too, and I don’t think you are showing that here (maybe you average those too?).
BUT I am getting beyond what can be done in Excel too. The statistical method you need is “Piecewise Regression”, and you let the best model fit “decide” where to split the data (Minimize SSE, or MLE), rather than doing it “by eyeball”. I’m still not sure this would be be better, but it is a less biased way to draw two lines as you are doing. The choice of what data to include will make a bigger difference, as you found with hobbit and Naledi.
I am swamped lately, so I can’t do this for you (not this month).