*sigh*

<?php
echo mb_detect_encoding("test?");
?>

use iconv -f utf-8 -t iso-8859-1 INFILE > OUTFILE
if "file OUTFILE" says utf-8.

Ignore the last comment. Do this:

Create two files with the following content, and name them test_iso1 and test_utf8:
<?php
echo mb_detect_encoding("test?");
?>

Make sure the encoding is correct:
$ file test_iso1
should return iso-8859-1
$ file test_utf8
should return utf-8

If they do not return the correct encoding, use iconv to convert them, e.g.
$ iconv -f utf-8 -t iso-8859-1 test_iso1 >test_iso1.fixed
or
$ iconv -f iso-8859-1 -t utf-8 test_utf8 >test_utf8.fixed

Now run each script. The test_iso1 script should return a type of iso1, the test_utf8 script should return a type of utf8.

Workaround: append an extra character to the end of the string, and then remove it(!)

It is not a bug, it is a specification.
You should use 'strict' mode in mb_detect_encoding() 
if you need to return correct result.

mb_detect_encoding() treat the string as byte-stream.
{0x61,0x63,0x63,0x65,0x6e,0x74,0x75,0xe9} is a correct
UTF-8 byte stream.
In this case, 0xe9 is treat as the first byte of
multibyte character. 

{0x61,0x63,0x63,0x65,0x6e,0x74,0x75,0xe9,0x65} is wrong
UTF-8 byte stream because 0xe965 is invalid byte sequence in
UTF-8.

If you need to remove the incomplete multibyte character from detection, please try to use 'strict' option like,
echo mb_detect_encoding($s1 , 'UTF-8, ISO-8859-1',true);

So to use my example from before, why do both
 $string = "test?"
in a utf-8 text file, and 
 $string = "test?"
in an iso-8859-1 file (converted using iconv) return "UTF-8" with mb_detect_encoding, even when strict is on?