Although you are technically correct, if you look at the data you'll recognise that for this narrow span of values, the logistic fit will be practically equivalent to the linear. Indeed, they perform the same in cross-validation. Here's what the logistic fit looks like: https://i.xkqr.org/logisfit.png
Author gave me the same reply. I just don't want to have to think about whether it's equivalent or not, why use a 2 param model that's strictly less appropriate even if the difference is small.
