I am wrong and I cannot figure out why !

I am working on matching a simple passive network to 50ohms at 2.4GHz using a board of FR4 three layers, first one just 8mils thick.
Initially Z=28+j63 and I wanted to go toward 50ohms. From Smith Chart the best guess would be : first serial cap of 1.8pF to go close to 28+j25 and then a shunt of 1.1pF to reach 50ohms.
Guess what: just adding a serial cap is moving me on Smith Chart as if I would add a shunt of 25nH. Is there anybody so kind to explain to me what is going on and what I have to care more in this type of matching ?
Thank you in advance.
Humane Society

It appears if you are working with the conjugate impedance. Make sure which one you are using and working towards the load or generator. If your circuit is like below it should work unless your components behaves strange at that frequency or you have a lot of stray capacitance. Zl = 28+j63 in diagram

Mr. E-design,
I am using the Smith.exe from Dellsperger Fritz ( free to download but very handy in use ). From that Smith Chart tool, just entering my starting point of impedance and choosing the working frequency which is 2.4GHz, I can figure out quickly which component I have to add. However, once I put my component on the board the new impedance I am measuring is far from the point I was expecting. At one point I was suspecting the FR4 thickness which is just 8mils, but I am not anymore sure on anything at this point.
I am in a complet dark.

HS

What size and type of capacitors are you using? Any stray capacitance to ground will have a dramatic effect, check for yourself. Maybe the solution below is better to absorb stray capacitance effects.

Mr E-design,
I really appreciate your help. Stray cap and inductance are an issue which I didn't see too much covered on this forum or some other books or tutorials. Beside that, hands on it - experience- is also important. I just discovered that the board I am working (using 0402 components) needs at least 5min to cool down and I can see how the impedance is moving while is cooling. Obviously the 8mils dielectric is a nightmare for design and tuning.
People who think that they can simulate that, are not honest.
We will stay in touch on this subject because I am going to be really bad with a lot of high position guys who don't want to share their experience in microwave and just watch other people talks.
Thanks
HS