The HTML file in question is Latin1 plus Latin1 Supplement
http://www.w3schools.com/charsets/re...supplement.asp, which is cp1252 less the characters in the 120 to 159 range. The displayed text of this HTML file, however, contains characters in that range, as well as two-byte utf-8 characters.
I had assumed that the "input character encoding" referred to the characters in the HTML file, not the characters displayed by the viewer or browser when interpreting that file. I thus tried both cp1252 and Latin1, neither of which produced correct output (in epub, mobi, or azw3 format). However setting the input to utf-8 does work. It would seem that a definition of this (to me) idiosyncratic use of "input character" be provided.