The possible pitch values equate to
0 = 1 Phrase (contiguous data)
1 = 2 Phrases (every other phrase)
2 = 4 Phrases
3 = 3 Phrases
3 is a special case. In the prototype blitter, this value was 8 phrases, but it was suggested that the value of 3 would be incredibly useful as it would allow two frame buffers to interleave with the z buffer data, so the blitter design was modified to change the encoding for 8 phrases to become 3 phrases.
But in reality,a lot of these weird thing in the jag were not well documented and nobody could help with them, so what we did was just TRY all 4 values and see what they did. If one value did what we needed, then we used it without worrying about why it worked or why the documentation was wrong. If it was an obvious typo in the docs, we would report it but nothing much happened to our reports. We didn't spend a lot of time trying to figure out why the documentation or examples were wrong.