As i promised - i maked new test on heroic dummy in IF.
I have 263 hit rating, FA 0\3
I have maded 2200 autoshots and pet all this time was attacking heroic dummy too.
Both of us never missed.

P.S. As im from Rus Realm - so i also have Rus version of WoW and Recount. But Dodge, parry and misses somehow haven't translated by author of the Recount addon anyway

http://img16.imagevenue.com/img.php?..._122_739lo.jpg -- Thats my autoshots
http://img137.imagevenue.com/img.php..._122_842lo.jpg -- pet's white attack
http://img224.imagevenue.com/img.php..._122_430lo.jpg -- pet's Claw ability
http://img245.imagevenue.com/img.php...4_122_82lo.jpg -- pet's Rake ability (dot ticking here)
http://img24.imagevenue.com/img.php?...122_1116lo.jpg -- pet's Rake ability

So, seems pet hit cap is also 8% as i said before.

So it seems we have 2 conflicting data.
I'll once again do some math thingies with them.
The question here would be which is more unlikely (or to translate to some stat terms, which is more standard deviation away from the norm).
First, getting 58 misses out of 2000 attacks with 5.95% hit, assuming 8% hit, has z score of 2.68.
Second, getting 0 misses out of 6243 attacks (3794 melee + 2071 claws + 378 rakes) with 8.02% hit, assuming 9% hit, has z score of -7.86.
Those z scores tells you how many standard deviation away from norm these tests are, i.e. farther away from norm they are, less likely they will occur, thus less likely the assumption is correct. Now, both these tests results are very unlikely to happen given each assumptions. Now, second test gave more unlikely result, so 9% hit cap assumption is more likely to be the wrong assumption.
Since there is too much data supporting 8% hit cap (all those WWS runs with no misses while not at 9% gives extremely small chance that 9% is hit cap), I'll just go with that for now, but these two tests supporting 9% hit cap is quite interesting (read statistical anomaly). 2.68 standard deviation away has less than 1% probability, but it's still far more likely to happen compared to other tests.
Only real conclusive result would be if someone missed with over 8% hit, but since that hasn't happened yet, so rest of data can only interpreted by some probability calculations.

Another thing would be if certain servers are bugged and not calculating hit chance correctly (or in this case, only few of them calculating correctly at 9%). This would account for these different test results with such unlikeliness.

I will be testing more tomorrow, as my pet got 7% miss with 0% hit from gear and 1/3 FA, roughly 6000 attacks.
The same happened to me, as expected, over roughly 3200 attacks.
I think that somehow, our hit from gear is overwriting hit from FA for our pets.
(think the screenshots are in the pet thread)

So more on that tomorrow, I'll be shooting thousands of arrows with 1, 2, 3, 4, 5, 6, 7 and 8% hit on gear.
To get somewhat more conclusive data.
I'll probably also send in a ticket to ask blizz about this and to kill time :P

I've already wrote that FA does not transfer to the pet.
Ive wrote ticket in game and GM said that there is no any bug, talent(FA) work as it must.
So, if its truth (i mean GM really knows how should work this talent and how it working now) - there is no point to spend 3 points in this talent as pet give us almost half of our damage (I mean Beast Master of course).

Just did Patchwerk with 5.76% + 2 FA (7.76%) and missed 1 Auto Shot. Pet had 3% miss. All according to Recount.

I just got new gloves so im gonna be using them for the rest of the night. 6.34% + 2 FA (8.34%) And checking if i miss for the rest of the night. I'm combat logging right now and will post the WWS later tonight.

EDIT: Heres the Wow Web Stats, i was able to get in after the Spider wing and Plague wings were already cleared.

Patchwerk was with 7.76% hit and everything after was with 8.34%. Only missed 1 time and that was on Patchwerk.

So just some update from a non-hunter.

10 man archavon, 25 man archavon, and 25 man pwerk with 8.03% hit. No misses over 800 or so swings so far. I don't really raid 25 mans (they needed my debuff for pwerk) so I won't have a chance to test till I do the 10 mans over the week. Would love to see more people testing out the 8% theory.

I'm moving through naxx still. We finished patchwerk and with 7.6% hit I had 0 misses on patchwerk gothik and razuvious with sub 1% miss rates on loatheb and loatheb.

I did a test on the IF heroic dummy:

Autoshot and pet melee attacks only

263 hit rating (8.02%)
0/3 Focused Aim

2000 shots (cheapest ammo!)
3354 pet attacks

zero misses for either

Screenshots:

Hunter stats
Pet stats

Heroic dummy would have a loser miss rate than a raid dummy, the mobs are only 82.

The heroic name is misleading. The Heroic Training Dummy has the skull icon or ?? as level.

Tested on Heroic Dummy.

186 Hit rating(5.67%) and 3/3 FA.

2000 Pet autoattacks.
3.6% Miss.

screenshot