does DNAmAge function assume chronological age?

[AnmolAPardeshi]

Horvath (2013) used DNAm data to define age such that

1. weighted average of cpgs + intercept = some number - but this is NOT the DNAm age and need reverse-calibration
2. so the reverse will be based on the definition of original given as (a) F(age)=log(age+1) - log (adult.age+1) IF age<=adult.age and (b) F(age) = (age-adult.age)/(adult.age+1) IF age>adult.age ; adult.age set to 20 by Horvath.
3. so even if the weighted average of cpgs with coefficients is calculated, the predicted age on chronological scale (ie DNAm age) will need that reverse transformation.
4. for reverse transformation, F(age) can be set to the weighted average, adult.age is set to 20 to by simple arithmetic, the "age" can be calculated which will be the DNAm age - but notice that which formula to chose from (a) or (b) WILL need knowledge of the sample's actual chronological age from covariates.
5. I manually implemented that reverse transformation for 1 adult sample and 1 infant sample and it worked - meaning the DNAmAge returned the expected values but when I repeat this on a non-example dataset, theres something weird going on.

Can you please help explain how without the knowledge of where the sample passes that "adult.age" threshold is the Horvath DNAmAge prediction being calculated?