A point worth investigating.

I'd encourage you to pull the data (as I do in this post) and assess the difference in Bayesian Information Criterion between a straight-line model and curve fit model!

If the curve fit has better information you may be correct.