Will doing the trend line off the data known to be complete (sounds like at least 3 weeks old) be good enough? If so, then you could filter that data out and run your trend line with a bit more confidence. Only other thing I can think that you might consider is to start analyzing what the data looks like that comes in late and if there is any consistency to the meaure values invovled that could allow for assumptions to be included into the current numbers.
Best bet is not including current data if trend line is important though. My thought at least.
Thank you very much. I will keep this in mind.