He forgets an important variable.
The density of gasoline also increases inversely with temperature. Therefore, at a lower temperature gasoline has more BTU per gallon.
Perhaps another contributor is increased losses at low temperatures due to the increase in lubricant viscocity.
It seems to me that he has neglected the factors of increased air density and fuel density on combustion efficiency (effect of mild supercharging increases power per stroke), and decreased road friction/stickiness, both tending to increase mileage. Before on-board computers, we used to have to retune our engines for best operation according to the weather.
The author clearly states that his analysis addresses changes in aerodynamic loss as a function of temperature only.
Many things affect mileage as a function of temperature, but they were not part of this guy’s work.
My wife’s hybrid Mercury Mariner drops from about 32-34 mpg on the highway to 24-26 mpg when temperatures drop below zero.