The caps should have a high Q but there is a point of little return for ones effort to go better.
If you use quality caps its generally not an issue at 12mhz, bigger deal at 1296mhz.
So for this its more a matter of using the right value for that set of crystals.
Paralleling gets better Q but for this it allows you to exactly hit the desired value.
as capacitors in the 30 to 100pf range are generally standard values only.
For example your could put 56pf and 33pf to get 89pf, or 47and 47 for 94.
I believe the optimum value centers around 91pf but, 1% caps are scarce and far from cheap
especially AVX and other MLCC types. So hand picking (with measurement) and paralleling
is the way to go. For a given set of crystals as they also vary so the "correct" capacitors is
in the range of 82 to 100.
Of course one could extract the crystals measure them and then calculate the exact value.
That's a lot of work and a simple change the values and try them is easier.
How wide should it be?
Other than my personal preference, a SSB filter is considered very tight at 2khz and wide
at 3khz so there is plenty of room to allow for anything in that range. Also I have two radios
that have adjustable digital bandwidth (IF DSP) that goes down to 200hz and up to 2800
so 2300 to 2500 is fairly normal sounding 2000 or less is really narrow sounding.
For radios where I've built the filter from the ground up I shoot for about 2400hz
a variation 100hz either way is hard to hear and still very good.
For CW nothing beats 500hz or so unless tuning then wider is easier.
Hope that helps.