I think I see what you are asking. If you design a bandpass filter with low loss L's and C's, it will have a nice low loss passband, and a good rejection band ONLY if the load is a broadband 50 ohm match. If the load is no 50 ohm somewhere in the rejection band, you may not get the desired ammount of dB's of rejection. This is because the bandpass is constructed of a ladder network of reactances that transform the load impedance all over the place internal to the filter in order to reflect any incident power in the reject band. If the load is not 50 ohms, then the internal impedances transformed in the filter are not what they should be, and performance suffers. An obvious example of this is to put two bandpass filters in series with a low loss transmission line connection of some length. Where ever that length is some multiple of λ/2 at a specific frequency in the reject band, then the stopband attenuation goes almost to 0 dB!
BUT, a SAW filter does not work that way. It uses acoustic methods to filter the signal, not impedance transformation methods. As such, I think it is relatively immune from load impedance.
That said, I never did actually test a SAW filter to see if the above is true.