I also thought about it again and took a closer look at the paper. Their sample includes dogs in the age of 0 to 15 years, so our calculated age is extrapolated. Probably the formula works in their range quite well, but at twice the age of their maximum age, the formula is perhaps no longer applicable. His age is probably somewhere between this result and the conventional formula.

According to this paper, the conversion is not linear but logarithmic, following the formula: human age = 16*ln(dog_age) + 31 Which brings his age to 85.8 human years. Here you can see this in a graphic, taken from the first figure of the publication.