While there seems to be insufficient proof that Lehman's tuning truly recreates anything used by J. S. Bach, it does adhere to the general goal that's implied in other historical well temperaments I've since read about: A gradual change in "coloration" of the tuning as you go around the circle of fifths, with the most "just" major third (i.e. flat of equal-tempered) occurring in the key of C and the most "Pyhagorean" major third (i.e. sharp of ET) at the other side of the circle at F♯. The width of the fifths in these temperaments appear to have been derived by trial and error to close the circle while achieving the desired thirds. Precision was of course limited by the mechanical precision of the piano and the capabilities of the human tuner.
I decided to develop a well temperament mathematically, using a cosine function to adjust the width of the fifths gradually. I used a spreadsheet to derive the thirds, and below are links (to Google Sheets) and a copy of the output, with two variations. The amount of stretch was chosen so that the widest fifths would be perfect (i.e. 3:2, about 701.955¢). For each of version, the flattest third (C-E) ends up about 7¢ sharp of just, and the sharpest third is about 20¢ sharp of just (but not quite Pythagorean).
Centered on C, with perfect fifths at C♯-G♯ and G♯-D♯:
Centered between C and G, with a perfect fifth at G♯-D♯:
Here are the tunings in Scala format if you would like to use them:
! Cosine C Well.scl
C-Centered Cosine Well
! Cosine CG Well.scl
C/G-Centered Cosine Well
Incidentally, these are rather nice-looking when viewed through Scala's temperament radar: