The best way to describe my tables is that I only allow second party material in my game, in that my tables tend to be so heavily house ruled that there is little point in using any documents but my house rules.
In general, I make no distinction between first party material (published by WotC for D&D example) and third party material (published under the OGL by Green Ronin for example). What matters to me is whether I find the material the superior implementation of the idea, or which material I find more inspiring. I have no idea really of 'canon'. The fact that it appears in a WotC book in no way means that it is favored. And in any event, there is a good chance that it will be heavily massaged by me before it arrives in the house rules.
Two examples might explain this well - Green Ronin's 'Holy Warrior' class and Green Ronin's 'Shaman' class.
Green Ronin's 'Holy Warrior' class is an attempt to fix the most salient problem with the 3.0e Paladin; namely, that the RAW WotC paladin is only a suitable champion for Deities that emphasize Justice and Righteousness as the key virtues which are laudable in a hero. Deities of a different character or moral philosophy we would expect to have champions of a very different character. This is a very old problem in D&D, as classes like 'The Elven Cavalier' and 'The Anti-Paladin' from 1e and 'The Blackgaurd' from 3e will demonstrate. However, from my perspective the 'Holy Warrior' as presented by Green Ronin still had some severe limitations: it required a specific write up for each deity, there was a lack of balance between concepts, some of the rules were klunky in application, and the class remained heavily front loaded. To resolve these problems, I wrote a homebrew class called 'The Champion' which is not as front loaded, is flexible enough to accommodate any deity without a need to generate custom content, and is IMO better balanced and less clunky. Is this '3rd party content'? Well, if this were an essay I was writing in college, I'd need to cite the original author to avoid plagerization, but I can sincerely state that not a single sentence remains from the original class write up.
Likewise, I really wanted to do something about the RAW Druid for similar reasons. The RAW Druid was tightly bound by flavor to Northern European myths and wasn't in my opinion either flexible or balanced. It was a tier 1 class that could outperform the fighter in combat at most levels, and at the same time compete with both the wizard as a general caster and with the cleric as a healer. That the Druid wasn't flexible in flavor was also something that was obvious from the proliferation of 'Shaman', 'Witch', 'Binder' and other occult magician classes. What I wanted was something much more in the flavor of the Green Ronin shaman, which could emulate much more precisely the concept of an animistic priest whose sorcery was based on appeals to spirits natural and unnatural. The problem was that the writer of the Green Ronin shaman was, for obvious reasons, writing the class to exist alongside the RAW Druid rather than being a replacement for it. Since I wanted the Shaman to replace the Druid completely, it had to do something the Green Ronin writer had pointedly not done, which was kill the Druid and take its stuff. This required a major rewrite of the Shaman's spell list so as to now encompass the Druid spell-list without becoming overly broad and competing directly with the wizard.
So do I allow 'Third Party Material'? I don't allow Paladins, Druids, Rangers, Monks or Barbarians! If you were to pick up a WotC book and say, "I want to take this.", chances are the answer would be "No." If you were to pick up a third party book, the odds of me saying "Yes." wouldn't change measurably. What I probably would do is suggest that the concept for the character could already be played within my rules as they existed, and that if the character could not be played ("I want to be a Were-Orca Minotaur Monk.") that either the lack of inclusion of the concept was deliberate or that if the idea worked for me as well we could work together to create some rules extensions that would support their idea.