The f-number marked on the lens is for the lens positioned one focal length from the negative, which would make an infinitely large print. As you move the lens away from the negative to make a finite size print, the lens becomes slower and the effective f-number becomes larger. This is the same problem known to larger format photographers as bellows extension. In large format photography, the effect becomes significant (approx 1/3 stop or more) when the image size is 1/8 lifesize or more. Of course, this is the typical range of enlargements when making a print so the effective f-number of enlarging lenses in use is normally larger than the marked value.
If you look up the formulas for exposures compenstation for bellows extension (closeup photography) in a book about large format photography you will find the correct forumlas that have been given on this and previous threads about changing print sizes, i.e., the formula using (magnification +1) and the formula using image distance. Both formulas account for the lens being focused and not being one focal length away from the image.
The two formulas are physical/mathematically equivalent (when the same lens is used for both print sizes) since image distance = focal length * (magnification + 1). T2/T1 = (m_2+1)^2 / (m_1+1)^2 = (d_i2/f)^2 / (d_i1/f)^2 = (d_i2)^2 / (d_i1)^2.
References: Applied Photographic Optics by Sidney Ray, p. 523; Lenses in Photography By Rudolf Kinglake, p. 99; Photography with Large Format Cameras by Kodak, p. 47; View Camera Technique (5th ed.) by Leslie Stroebel, p. 182.
A further detail brought out by Sidney Ray is that the pupil magnification factor should be included in the forumula. But since most enlarging lenses are fairly symmetric, this can probably be neglected.
Another source that was offered directly for enlarging was the "Enlarging Dial" of the Kodak Darkroom Dataguide. This is a circular slide rule-type calculator that can be used to calculate changes in exposure time with print magnification. If you "reverse engineer" it, you will find that it very closely but not perfectly follows the (m+1) squared formula. The times that it gives, for cases of at least one of the magnifications being small, are far off the purported magnification squared formula.