Time sync of the Extend: Loop function

[BrianP63]

Firstly I would like to say that I am completely new to SoundThread and CDP. I only found out about it recently when watching a MylarMelodies YouTube video, that gave me yet another rabbit hole for my neurodiverse brain to spend hours disappearing down. So if I am asking in the wrong place or asking a question that has already been answered I apologise. I would also say I appreciate Jonathan's efforts in developing Sound Thread.
To try to get to grips with what SoundThread and CDP are doing I decided to pick one of the function and study it in some detail. Having watch one of Jonathan's videos I decided to look at the Extend: Loop function. After listening to the effect on a few random audio samples I decided to apply it to a drum loop. It give a very interesting result but I noticed that later in the output file the beats were no longer in sync with the BPM of the original file. To test this further I created a 4 bar 100BPM file containing only kick drums on the beats. This test file was a 44.1Khz 24bit WAV file output from GarageBand, shortened to be exactly 9600ms in Amadeus Lite. I don't think the overall length of the input file should be significant, but I did this just to be sure. I tried various iterations of the function using step and loop lengths in multiples of 600ms, but always found the output file would drift away from 100 BPM within the first few bars.
I found varying the loop length would resolve the issue but a bit of experimentation was required to achieve this. Using a loop length of 4854ms and step length of 600ms produced a file of 7 minutes 7 seconds, which stayed at 100BPM for the whole file. And a loop length of 4908ms and step length 1200ms, produced a 3 minute 36 second file which did the same. I tried keeping the loop length at 4800ms and varying the step length but couldn't make that work. The output files were checked by importing them in to GarageBand.
I may have misunderstood what this function is meant to do, but my understanding of what I have read and the description given in Jonathan's video is that the output should maintain the BPM of the input if loop and step lengths are in proportion to the BPM. I realise there may be may parameters at play, but the CPM site description of the function includes the following line, "The scat parameter randomises the length of step (within a small range), producing a less mechanical result." There does not appear to be a scat parameter available in SoundThread, although the results I saw would not indicate a scatter issue, it does appear to be an offset linked to the loop length. There is also a splice length parameter (wsplen), described in CDP, which has a default of 25ms. I think this could be the issue but I does not exactly correlate to the increases I was making to the loop time to maintain the BPM.

If Jonathan or anyone else can enlighten me on what is happening here I would appreciate it. I never rule out the possibility I am being dumb and have missed something very obvious.