All my datasets met the same problem that the resolution numbers look good but the true resolution seem to be much lower.
I found that all of the unmasked FSC curves plummeting to 0.4-0.6 like the following pic. (~540,000 particles)
Other refined curves also dropped around same resolution range.
They all look like basin-shaped at the same resolution range.
I am wondering is this feature normal?
Is there anything can lead to this kind of FSC curve?
What can I do to improve it?
Yes absolutely - particularly with 540k particles remaining, plenty of room for improvements! Just wouldn’t worry about unmasked FSC being worse than masked, was what I meant