"So", in this context, is equivalent to "Uhm" or "Like" or a dozen other null noises people use when they're nervous. Trying to analyze my own use of one of these (I won't say which), it seems to actually be a mechanism to make sure I'm actually producing voice at appropriate volume level before I start modulating it. I couldn't tell you why I feel I need that in some situations, but apparently I do.
So (and here I used the word more appropriately), I'd say that the real problem here is that the transcriptionist should have edited out the leading "So", just as they would "Uhm", but chose not to
As verbal ticks go, this is a minor one. Worked with someone who had one that ran about 10 seconds, and was repeated every few minutes. We quickly learned to either ignore it, or to make internal bets about how many times it would come out in any given presentation..