Misinformation?

Well, I'm trying to get this sensor up and running, but I'm really strugling with a lot of misinformation.

First of all my evaluation board doesn't match the documentation at all. Look here:
viewtopic.php?f=3&t=875

Anyway, I got it working now by simply guessing and assuming, and I'm now able to communicate with the device over SPI.

Next I have trouble understanding the FIFO behaviour, which I honestly find is maybe not explained particulary well in the documentation.
Like, can I for example read X bytes from FIFO_R_W in one transaction (which it seems like), or do I actually need to setup the address for every 1 or 2 bytes.

Next, the FIFO count states that the value read is related to the number of samples, however I'm pretty sure that I'm reading the number of bytes?
When the FIFO is full, I read 0x0400 which is 1024 bytes.

I've set the sample rate to 1KHz and the divider to 200 producing 5 samples per second.
After initializing and waiting 500ms I read 0x0012 when only enabling the Gyro FIFO (XYZ). This corresponds to 3 sets of 6 byte XYZ samples (Gx+GY+GZ).
Enabling the acceleration FIFO as well, I read 0x24 which corresponds to 3 sets of samples each of 12 bytes (Gx+GY+GZ+AX+AY+AZ).

According to this, I should read the number of "samples" and not "bytes" in the FIFO:
viewtopic.php?f=3&t=604

So from my point of view, the board and registers are not mathing the documentation at all.
Have I got a wrong evaluation board, with an early MPU6000 device - like maybe a pre-release?