CST： Different results using 2009 and 2010 versions?

HI, Last year I did a project where I mainly used CST Studio design for S-parameter analysis of different components, then I used the 2009 version.
This year I'm working on a larger project where I use some of the same components, and I use the 2010 version.
I just discovered something strange. I do the same simulations, just using an S-parameter job in studio design to check the same components, but the results differ. Not by much, but when I export to touchstone, and import in matlab to create figures where ie I compare S11 and S22 of the same component, they are not the same relative to each other as my old results.
Have they changed something in the solving routine? Or is it more probable that I am doing something differently? I to be honest, I cant really remember the settings I used last time, but they were not very advanced. I've tried changing some settings, ie. port impedance controlled by components, or setting it default to 50 Ohm. I've tried calculation for real/imaginary results and for magnitude (dB) results. Either way, the results differ from the ones I got in 2009.
I am confused

CST 2009 gives more accurate result as mentioned by my lecturer but i am not sure why...

How much do the results differ ? Are we talking about 0,5 dB or 5 dB ?
But you should be lucky that your models are still working... I've got models with multipin waveports which worked in 2009 and also up to 2010 with Servicepack 2. But after the installation of SP3 and SP4 they aren't working any longer ! CST Support confirms the problem, saying that version 2011 won't show this bug...
So maybe you should also contact the CST support ?!?!

Hi, CST 2009 gives more accurate results all over? We are here talking about simple S-parameter analysis by loading in TOUCHSTONE blocks in design studio, attaching ports and then run the solver.
Hi, as per above, maybe I should clarify. Its probably the simplest thing you can do in CST. Importing touchstone files (s-parameters) in to design studio as touchstone blocks. Then connecting ports and run the solver. This lets you connect one or more devices to see how the transmission properties change etc.
So, such a simple simulation. More like a recalculation and plotting routine, how can it produce different results from one edition to another? I mean, for only ONE device, I could just as well load the S-parameter file into matlab and make the plots as I want by simple calculations. ( I dont have a license for the RF - tool in matlab, so I cant do a simulation of several devices. )
There are two cases where I get different results. The first is when I simulate a SPDT switch. ONE simple file, and I get get different results for S11 and S22. We are not talking VERY much, but where the curves had a slight intersection before, they now only almost brush each other in the same range of frequencies.
The other case is a cascade of two and three LNAs. Here the old results showed that S11 worsened just a bit, like 1 dB difference, between two and three LNAs. In the new edition they are identical.
Strange indeed.