Is astigmatic CTF correction worth it?

I have a question about including astigmatism in the CTF correction of single particle reconstructions. Some people recommend not allowing astigmatism and just using an average value for the defocus in the calculations. The reason is that, unless the astigmatism is quite large, the additional defocus value and astigmatic angle that need to be determined add more error to the calculation rather than help restore high-resolution detail. Does anybody have experience with a data set that goes to high resolution?

I used our refined data set of rotavirus DLP to calculate a reconstruction that uses average defocus values instead of astigmatic defocus to do the CTF correction (with Frealign). The data set had been refined with data up to 9 Å resolution. Therefore, the FSC at around 5 Å resolution should show little or no bias. The two attached FSC curves (as calculated by Frealign) compare the reconstruction calculated with astigmatism (atigmatism.txt) and without astigmatism (no_astigmatism.txt). The latter FSC curve shows a clear loss of resolution. The astigmatism in the data set has a standard deviation of about 1500 Å. For this amount of astigmatism it is therefore clearly important to include astigmatism in the CTF correction.