This bit of code is intended to remove h-cards from implicit feeds of h-* when parsing:

It appears the logic doesn't work quite right if the h-card has no url property. In this example, the parsed mf2 is (snipped out parts not needed for this example):

"items": [
    {
        "type": [
            "h-entry"
        ],
        "properties": {
            "url": [],
            "published": [],
            "author": [],
            "like-of": []
        },
        "lang": "en"
    },
    {
        "type": [
            "h-card"
        ],
        "properties": {
            "name": [
                "RSS \u00b7 Webmentions"
            ]
        },
        "lang": "en"
    }
],

This is a bit more verbose in part to make it easier to follow, but I think the array_filter should be something like:

$tmpmf2 = array_filter($items, function($item) use($url) {
  if (!in_array('h-card', $item['type'])) {
    // not an h-card; keep it in the result array
    return true;
  }

  if (
    isset($item['properties']['url'][0])
    && $item['properties']['url'][0] == $url
  ) {
    // h-card's u-url matches page URL; keep it in the result array
    return true;
  }

  // h-card u-url does *not* match page URL (or it has no u-url); remove it from result array
  return false;
});

I haven't written a unit test for this yet, otherwise I would make it a PR. It did work in my local testing with the specific example link above.

I might be able to make a PR in the near future, or if someone else wants to take it on, feel free!