smoothing data with 'rloess'

I have a spO2 (oxygen saturation) signal. it contains round values around the 95%. But there is a problem, there are artefacts, thus outlieres. So i need to filter the signal without taking the outliers in the moving mean filter. So i used a 'rloess' filter. but if i compare the raw data with the filtered data, there are no differences. But the boxplot tells me there are a lot outliers. Has someone the solution?

[jonas]

Show us how you call whatever function you used to filter with. Perhaps you used a very small span/window?