with Divider2, VCO's phase noise does not improve 6dbc/Hz

Hi all,
I use vco+div2 as my LO generator.
here is the simulation results,
without div2, -105@100k, -135@3M
with div2, -110@100k, -140@3M (just 5dbc/Hz improved)
From the noise contribution, it can be seen that the devices of div2 contribute a lot(20%), especially the bias of div2.
Also, my vco output peak-to-peak swing is around 1v(single end) without any load, but turn to 700mv with div2 as loading.
I simulated the phase noise of the div2, and its PN from 100k to 10M is about -150dBc/Hz. so I think div2 itself should not contribute
so much noise.
Does anybody know what's wrong with this?

Thanks in advance.