As Dolf, PA1RW suggested: have you checked the phase noise of your 25 MHz source?
Absolute accuracy and phase noise/jitter are two entirely separate issues. Phase noise level will increase as a function of "multiplication" in the PLL process, which in your case sounds like 390 times what you have at 25 MHz. The external reference not only has to be reasonably accurate and long term stable, but also short time stable, including low DDS spur level close to the carrier (in the order of Hz).
As examples, the 25 MHz TCXO on the left is only just barely useable and the 25.787179 MHz one on the right is totally useless at their 390th harmonic:
https://73.fi/oh2aue/tcxo_ultr…lden_oscillator_clock.jpg
Here is a very simple spectrum measurement between the 25.000000 MHz TXCO signal, the 25.787179 MHz signal and only as an example, the 25 MHz signal from my R& SDX, with all frequencies overlayed for comparison:
https://73.fi/oh2aue/tcxo_vs_rs_smx_oh2aue.png
Already in this measurement you can begin to see there are issues.
And this is what a received, clean 10 GHz signal (my 10369 MHz beacon) looks like when using either the SMX of the 25.787179 MHz TCXO as reference for a modified PLL LNB:
https://73.fi/oh2aue/10GHz_SNR…TCXO_vs_RS_SMX_oh2aue.png
You cannot detect any kind of 10 GHz carrier with the TCXO programmed for 25.787179 MHz, there is just a phase noise hump left !!!
The SMX LO is not exactly clean any more, but at least it is not limiting my SNR on QO-100 reception.
For the TCXO programmed for 25.000000 MHz, the 10 MHz spectrum might be only barely useable as I mentioned earlier: the 10 GHz signal to noise ratio will always be limited to about 15dB due to "multiplied" phase noise, even when it could be 20 - 25 dB. Sorry, no plot for this, but you can imagine. And no, I am not using either of these TCXO's, they are both in my reference phase noise collection 
For microwave narrow band operation one needs (at least) accuracy, long term stability, short term stability and low close-in spur level. Well designed, ovenized crystal oscillators are very difficult to beat and if it is phase locked (with a very slow loop) to something more *accurate* like GNSS/GPS, Rubidium, Cesium (or even a Hydrogen MASER), then you just may have reached Nirvana 
/michael, oh2aue
P.S. and of course this multiplied phase noise plays havoc with (phase modulated) DVB reception.