Here's what I've got so far.
On your screenshot with the parts breakdown, the value that is sometimes 0x13 and sometimes 0x03 are the command which on the intensity set commands indicates how to interpret the channel fields. 0x13 indicates that it is a 2-byte bitmask of channels. A value of 0x03 is what we currently use in xLights and FPP and that indicates that the field actually contains the actual zero-based channel number OR-ed with 0x80. That must be how they send more than 16 channels to the 24-channel controller, they could use 0x03 with an actual channel number 0-23 or possibly there is a different code other than 0x13 which has 3 bitmask fields for the channel numbers.
For the CCR testing, here are some observations:
CCR all white for 3 seconds
00 02 51 49 3F
00 02 51 08 FF FF
00 02 51 07 FF FF
00 02 51 06 FF FF
00 02 51 05 FF FF
00 02 51 04 FF FF
00 02 51 03 FF FF
00 02 51 02 FF FF
00 02 51 01 FF FF
00 02 11 FF FF 00 00
The "00 02 51" lines are setting the channels to apply the value to. Were you using all 50 pixels or just 45? On each of the first nine lines, it looks like we are addressing a 'bank' of 16 channels, going from bank 9 down to bank 1. The first line might seem to indicate that we are only looking at a bank of 8 channels. This would add up to 135 channels or 45 pixels. For banks 1-8, all channel bits are turned on via the 0xFF 0xFF. The last command is the "all on" command as noted at the top of the doc.
The second CCR sample seems to agree with the analysis above.
First 10 channels red
00 02 51 C9
00 02 51 C8
00 02 51 C7
00 02 51 C6
00 02 51 C5
00 02 51 C4
00 02 51 C3
00 02 51 C2
00 02 51 01 24 09
00 02 11 49 92 00 00
In this, the "00 02 51" command is sent with a "C9", "C8", etc. which appear to be clearing the bits on those banks of channels. Then, it sends a "00 02 51 01 24 09" which would set the red channels 3,6,9,12 since the two bytes are reversed when looking at the 16-channel bitmap. It looks like when you said you wanted 10 channels, it actually did 12 since it sees them as RGB triplets. I'm uncertain what the "00 02 11 49 92 00 00" is at the end.