When you need to convert from HTML-ENTITIES, but your UTF-8 string is partially broken (not all chars in UTF-8) - in this case passing string to mb_convert_encoding($string, 'UTF-8', 'HTML-ENTITIES'); - corrupts chars in string even more. In this case you need to replace html entities gradually to preserve character good encoding. I wrote such closure for this job :
<?php
$decode_entities = function($string) {
        preg_match_all("/&#?\w+;/", $string, $entities, PREG_SET_ORDER);
        $entities = array_unique(array_column($entities, 0));
        foreach ($entities as $entity) {
            $decoded = mb_convert_encoding($entity, 'UTF-8', 'HTML-ENTITIES');
            $string = str_replace($entity, $decoded, $string);
        }
        return $string;
    };
?>