I have a D flip flop working well when it is simulated alone for all corners.

But when I add the flip flop in my modulator(continuous time DSM),  the flip flop drop a bit sometimes.

Both testbenches use the same errpreset as "moderate" and timing of data and clock is almost same.

For accuracy, I changed the errpreset to "conservative" and the flipp flop in the modulator works well.

Here, I wonder why the flip flop behaves differently for the same errpreset of "moderate".

The only difference is that one testbench(flip flop alone) is simple and the other(the modulator) is complex.

Does spectre simulate the flip flop roughly so that it drops a bit, for "modereate" errpreset when the flip flop resides in a complex circuit?

Thank you in advance.