[I'm the Enchant maintainer and author of the Enchant support in ispell.el, so it's All My Fault :) Sorry!]
I cannot reproduce the problem with numbers being detected as misspellings from the command line with enchant-2 2.2.8 or 2.2.10. (As all the digits are in WORDCHARS, it is expected that numbers should pass spell-checking.)
I looked into the matter of "-B". It would be quite tiresome to have to teach ispell.el which spelling checkers accept -B. I notice that hunspell does not support -B, but it does accept it; indeed, it seems to happily accept and discard any unknown option. I think Enchant should silently ignore -B, as it does with -m to work with Emacs.
I agree that the en_US dictionary should contain ' in its WORDCHARS. I found another, recent, Debian bug about this, this time on the correct package (the dictionaries):
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=964257 I have merged the two bug reports in the hopes that this might help!
Note that quotation marks work correctly in the Debian en_GB dictionary, though there is another problem there: WORDCHARS contains only the curly quote. This is because hunspell contains the following special rule: if either the straight or curly single right quote is listed in WORDCHARS, then the other is also allowed. I have implemented this logic in Enchant 2.2.10 in its hunspell backend (recently released).
I found a couple of minor bugs in enchant: first, it gives an unhelpful error message when asked for a language it doesn't have; and secondly, at least for me `enchant-lsmod-2 -list-dicts` does not list the en_US dictionary even though it is installed. I will look into those, as the second in particular affects Emacs.
I agree with the patch to ispell.el, in all three points, though for me the invocation of enchant-lsmod-2 does not fail with the extra argument (this will be why I didn't notice the problem when I originally wrote the code). I am happy to install this patch if Eli agrees.
--