time recompose and negative lower bound

[McCompexpert]

I have a statistical question:
In your cran downloads time series after time_recompose we see the grey area for upper and lower bounds. As the number of downloads can never be negative, why the lower bounds do?

Thus, it could happen, that the number of downloads is zero but still not an outlier because the lower bound is negative!
In general - how to handle this? Maybe log transformations of observed time series, do the analysis and then re-transformation back?

[McCompexpert]

Any pointers, please?