I can't reproduce your problem. -- provide sample data. (NT)
I think it is possible for a spline fit (the smoothed line option) to generate values below 0, and, for that matter, above some maximum limit.
In fact, if you open the graph wizard and select the scatter plot option, you'll see a group of graph types for X-Y graphs. One of these is a smoothed line graph, and even in their example, the smoothed line dips below the minimum value of the data to which it is fit.
I've got a call in to a math friend of mine who specializes in this sort of work, I'll let you know what he says- if I can figure out what he's saying!
Splines and Tigers and Beziers, oh my!
This is the message I got back from Dr. Alfeld regarding splined/smoothed data falling outside the raw data to which it was fit.
"Eric,
That depends on the precise scheme that's being used. For many, if not
most, schemes the extreme values of the interpolant and approximant
will be outside the range determined by the extreme values of the
data.
Sincerely,
Peter Alfeld"