Your anonymous function takes a "snapshot" of the variable myStruct as it exists when the anonymous function is taken. The only way to update that "snapshot" is to recreate the anonymous function. You can see this using the debugging tool functions. As its documentation page states, don't try using this programmatically. It's intended to debugging and diagnosing problems.
myStruct.A = 3;
myStruct.B = 7;
myStruct.myFunc = @(x) myStruct.A*x + myStruct.B;
data = functions(myStruct.myFunc)
data.workspace{1}
data.workspace{1}.myStruct
Probably the easiest option, since you say you don't want to define an anonymous function with 27 inputs (and I wouldn't want to define such a function either) is to define one with two inputs.
myStruct.myFunc = @(x, myStruct) myStruct.A*x + myStruct.B;
myStruct.A = 3;
myStruct.B = 7;
myStruct.myFunc(1, myStruct) % 10
myStruct2 = myStruct;
myStruct2.B = 6;
myStruct.myFunc(1, myStruct2) % 9
This has an added benefit of allowing you to pass the same struct array into multiple functions. Each one uses the pieces of data it wants and ignores those pieces it doesn't.
myStruct2.C = 46;
myStruct2.myFunc2 = @(x, myStruct) myStruct.C-x./myStruct.A;
myStruct2.myFunc2(12, myStruct2) % ignores myStruct2.B entirely
