Plot two data series on the graph, and add the trendline to the series
that does not have anomalies
Jerry's approach has the added benefit that it will allow you to plot the
"anomalous" data points with a different color/marker style and label them on
the chart key as being excluded from the trend line.
This makes them obvious on the chart and prevents people from thinking you've
cooked the data -- or at least lets them understand how you cooked it.
Bill -- (Remove KILLSPAM from my address to use it)