Apostrophe vs its html hexadecimal notation conflict - php

Question

I'm writing a little class to minify JavaScript with PHP. I have the following problematic code in my class:
private function test_opener($str, $i) {
if(ord($str[$i]) === 34 or ord($str[$i]) === 39)
{
if($this->_is_string_opened)
{
if($this->_string_opener === $str[$i] and ! $this->is_escaped($str, $i))
{
$this->_is_string_opened = false;
$this->_string_opener = null;
}
}
else
{
$this->_is_string_opened = true;
$this->_string_opener = $str[$i];
}
}
}
My class loops through each character in the file. The function above detects string opening/closing haracters (' and "). 0x34 and 0x39 are the character codes for " and ', respectively. If one of these characters is detected, a is_string_opened will be flipped to true if this is the first character opening the strong, or false if the character closes the string.
Now, my code breaks when I try to minify the following JavaScript (which is taken from the source of Underscore.js):
var entityMap = {
escape: {
'&': '&',
'<': '<',
'>': '>',
'"': '"',
"'": ''' // Here be dragons
}
};
entityMap.unescape = _.invert(entityMap.escape);
So what's append when the parser reach ''' : The first ' switch the _is_string_open to true. ', which is the HTML hexadecimal entity notation for ', turn it off, and the last ' turn it on again. So the rest of the code is interpreted as text until the next ', which is obviously messing the file parsing process.
I don't understand this PHP behavior. The character code of &;#x27; isn't even 39, it's 38. I ran the code on PHP 5.5.9. The encoding is UTF-8 and the content come directly from POST, but i try to add a htmlentities() to escape this kind of problematic character, nothing changed.
Edit : The data origin (a Controller getting post data)
$js = $_POST['javascript_content'] ?: null;
if($js)
{
$output_js = Jsmin::forge($js)
->min()
->join()
->get();
}