What is the difference between LOESS and LOWESS? From Wikipedia I can only see that LOESS is a generalization of LOWESS. Do they have slightly different parameters?
I think it is important to distinguish between methods and their implementations in software. The main difference with respect to the first is that lowess allows only one predictor, whereas loess can be used to smooth multivariate data into a kind of surface. It also gives you confidence intervals. In these senses, loess is a generalization. While the default for lowess is to use the tricube weighting, loess carries out an unweighted fit by default.
Now for the implementation. In some software, lowess uses a linear polynomial, while loess uses a quadratic polynomial (though you can alter that). The defaults and shortcuts that the algorithms use are often quite different, so that it is hard to get the univariate outputs to match exactly. On the other hand, I am not aware of a case where the choice between the two made a substantive difference.
Specifically for R, the difference is small. There is a very detailed explanation here: https://support.bioconductor.org/p/2323/
But notice that lowess() in R outputs data list while loess() outputs the model which can be input into predict().