OK, I have taken a closer look at the analysis on the web page at this URL:
http://www.amwindow.org/tech/htm/diodedistortion.htm
and it is more screwed up than I thought.

snip a vastly complex and incomprehensible disputation of the
largely incomprehensible text and formulae at
http://www.amwindow.org/tech/htm/diodedistortion.htm

About all we want is low distortion detection, and it matters noe that we
cannot follow all this mathematical analysis.

There is no mention of the output voltages measured with respect to
the % of modulation.

But anyway, a table at the conclusion of the article
gives the thd at various %m, :-

Modulation
Index (%) THD
(%)
10 1.02
25 0.08
50 0.32
100 2.0
150 6.3

Table 2 - Measured Total
Harmonic Distortion Versus
Modulation Index

But we dunno what the output voltages are, and no doubt the
thd results would be very different if the output voltage was 10v instead of say 1v
at 10% modulation, especially with a solid state diode.

From the test circuit shown, there is no bias current flow in the diode to keep it
turned on even
without an RF signal to demodulate.
This would also reduce thd.

Nobody needs to know math involved with diode detectors
to get much lower thd than is realised in most old fashioned and attrocious tube
detector stages in
conventional AM radios.

Patrick Turner