> I have tried to leave the accented characters in and to use iso-8859-1, 

If you are typing the characters in iso-8859-1 then you can just type
them directly from your keyboard and then use
<?xml version="1.0" encoding="iso-8859-1"?>'
In theory an xml parser is not forced to accept iso-8859-1" encoding but
in practice it will.

You are using
<?xml version="1.0" encoding="UTF-8"?>
which any XML parser will accept but in that case you would want to type
your characters using utf8 encoding. (How to do this would depend on the
system that you are using)

If you want to keep your files as being essentially ascii and use
character references to access non ascii characters then you can use
either UTF-8 or iso-8859-1 as they are the same in the ascii range.

but then you want to use & # 2 2 4 ;

(without spaces)

> I have tried to use &agrave; entity, but & is a reserved character.
entities must be defined before use, so if you have not specified a DTD
that defines agrave to be some character you will get an undefined
entity reference error.

> I have tried to use &amp;agrave; 

By quoting the & you are just specifying the string"& a g r a v e ;" You
are explictly specifying that this is _not_ a reference to teh agrave
entity.

>  or &amp;#223; 

similarly this is just the string "& # 2 2 3 ;" not a character
reference.
You want to use & not &amp; so that you get a character reference 
(but 223 is sz you want 224 for a grave)

David