The Document system identifier is null if you provide your own content handler.

[ndw]

I think the problem is that in Driver.java:

    public void tokenize(InputSource is, int bufferSize)
            throws SAXException, IOException {
        if (is == null) {
            throw new IllegalArgumentException("InputSource was null.");
        }
        tokenizer.start();
        confidence = Confidence.TENTATIVE;
        rewindableInputStream = null;
        tokenizer.initLocation(is.getPublicId(), is.getSystemId());
        ...

The call to tokenizer.start() initializes the HTML Document object with a null Locator. I think it would work correctly if tokenizer.initLocation() preceded tokenizer.start().