In a inductively degenerated common source LNA, I am assuming you are trying  to move the noise circles closer to the S11 50 ohm s parameter. Practically the top number one consideration what is often overlooked  is not considered is the Q of the inductors of a inductively degenerated common source LNA. This will effect your impedance matching. The second priority is picking a input blocking capacitor that is low loss and resonance free at you operating frequency. A blocking capacitor Q and loss will effect performance of the LNA. The addition of input-blocking capacitor should not affect the performance of impedance matching, however its Q and loss will add to noise figure. The bias circuits should be modeled, and a stability K factor analyzed for your LNA. If there is a oscillation even at a lower or higher frequency outside your desired LNA frequency your noise figure will be much higher. Be sure your LNA is stable you have modeled and measured your LNA.      

In general, large value of blocking capacitance is considered to avoid distortion of the signal at particular frequency of interest. nF range would be sufficient enough to keep |Xc| less than 1 Ohm if I am interested in 2.4GHz band signal. As Timothy said, Q factor need to be considered.

As said by previous friends, in general blocking capacitor should not affect the matching performance if it is large enough.  However, in some designs, somehow small capacitor may be purposely chosen to improve input impedance matching, particularly when the inductive degeneration is applied to improve the linearity of the LNA.

The Input blocking Cap is made larger so as to not affect the input resonance, by being a low impedance at our frequency of interest. As a consequence we could expect it to resonate at a lower frequency. In that case we would definitely need a BPF prior to LNA in the Rx circuit, correct?

Besides can someone explain how a smaller cap is purposefully chosen to improve input matching?