The higher the order of the filter, the more critical the displacement of each element becomes from calculated bandwidth, which ultimately does have a cascade effect on the final response of the whole filter. What you can do is to perform components association series/parallel in order to try to tune up to the desired value, although tolerance remains the determining factor on deviation from original specs. Monte Carlo analysis is quite useful in determining the components with the greatest impact on performance degrade. Keep in mind that at RF frequencies, layout matters, so that some of the above small values ( 0.2pF ! ) do not even exist to sell, or are in the same magnitude order of the parasitic elements of the components/board.