Typically when using FM (or in actuality, phase modulation just like the DX7 and so on) you want to reset the phase of the oscillators on each note on/trigger. This causes a sudden change in the waveform, resulting in a click.
However, an Amp Attack of the default value will remove this sharp click.
Being able to set the Attack to such low times can be effective for sound design, for example if you want a sharp transient at the beginning of a bass drum or hihat.
But besides that, you need to consider whether the oscillators are resetting or free running, whether the amp envelope is resetting or not and if the voices are stealing in a way that causes clicking.
With voice stealing, a click can typically occur because the voice that is stolen goes from sound A to sound B, sound B having parameters that are drastically different - resulting in a click. A typical culprit is the filter mode for example.
This is of course not an issue with many other FM synths as they are not behaving like a groovebox.
This can however be mitigated by simply deciding the amount of voices per track, ensuring that no voice stealing actually occurs, or just limit the voices of the most drastically different track.
I’ll try to write a more in-depth guide/explanation when I have the time, but I hope that you can see that the engine itself isn’t inherently ‘clicky’ or somehow ‘wrong’, it’s just that when you have this versatile system with tracks sharing voices with patches that can be very different these things are near impossible to avoid without having lookahead. (Which is a luxury hardware does not have)