My question is where do I have to place an impedance matching circuit?
P1, P2 or both?
Your attention should gravitate toward the input side (P1). As @Charly says, assume that output port (P2) is well-matched, and appears as a resistive 50 ohm load. This assumption may be dangerous in the case of cheap SDR receivers, since their designers are far less rigorous than designers of spectrum analyzers in achieving a well-defined input impedance.
Adding components to achieve a match at this critical low-noise point (where losses can impact signal-to-noise ratio) should be done reluctantly, so a designer might modify the filter's input side - you should be able to achieve a match by adding only one extra part. If you add lumped-component matching outside this filter, two extra parts are needed.
One might regard capacitance as more-easily variable than inductance: both C1 and C2 might be made variable. If L1 is a hand-wound copper air-core coil, it might be regarded as variable too, instead of making C2 variable.
This network can accommodate an input resistance up to about 1200 ohms. A pure 1200 ohm resistance allows one to merge C1 and C2 together:
simulate this circuit – Schematic created using CircuitLab
For input resistances lower than 1200 ohms, the input side of this network could be modified by making both C1 and C2 variable:
simulate this circuit
One has a problem if input reactance cannot be accommodated by the tuning range of variables C1, C2. In this case, a fixed coil or capacitor would be added to tune out the excess reactance.
If input impedance falls below (roughly) 40 ohms, try using a 10pf variable capacitor for C1...it can accommodate an input resistance as low as 10 ohms.
BTW: the filter center frequency as shown is a bit low for weather satellite - its centre is 134.5 MHz. Both L1 and L2 should be reduced a bit.