Consider the relationship between the present day 207/206 and time:
207/206 = (1/137.8...)*[exp(l235*t) - 1]/[exp(l238*t) - 1]
where l235 and l238 are the decay constants for U235 and 238.
Making a plot from this equation looks something like below:
You can see that when the 207/206 value goes below about 0.045, the "time" becomes negative, and as the ratio gets even lower the time rapidly becomes more negative.
Obviously < 0 time does not make sense in this context, but since iolite calculates an age for each time-slice of data, noisy and/or very low 207/206 measurements can have very large negative ages when the ratio dips down.
Now the question is what to do about it: If we were to nullify the time-slices that result in t < 0, this would artificially bump up the mean time. On the other hand, keeping them in drags it down. I think the best option is to calculate the age from the mean of the ratio and use the time-resolved "ages" in iolite as guides when making selections or for imaging. Calculating the ages this way could be done via a python-based exporter or third party software like Isoplot. We are also planning to add a feature to optionally calculate ages in iolite as "associated" results (i.e. not time-resolved and using the mean ratios) in an upcoming release.
All the best,