I have little experience in this area, but I believe this represents the relevant conventional wisdom:

In one long/legato note from a string or horn or wind, each note usually has some kind of movement/expression dynamic performance in it, as string/horn/wind players do not hold their notes at a constant volume throughout each note.

Samples generally do not record a lot of that movement/performance, and instead record a relatively flat note, which then obligates the user to put back in those dynamics. The way most (e.g. Spitfire) libraries are designed is to assign long note dynamics/volume to a slider or wheel (or foot/breath controller) which the user is expected to move continuously while playing, to mimic what a normal string/horn/wind player would do.

Since key velocity gives you only one volume level for the entire note, the key velocity control paradigm is fundamentally incompatible with most long notes from most orchestral instruments, and (whether technically possible or not) it is more or less incompatible with the design of most orchestral sample libraries (such as the Spitfire orchestral libraries); trying to play such libraries within the key-velocity-controls-dynamics paradigm will handicap the user.

Moreover, many/most libraries (including those from Spitfire) use velocity during legato lines for another purpose like legato type/speed control, so trying to shoehorn velocity control of dynamics into the dynamics of legato notes will often handicap the user in multiple dimensions.

Odds are, it is best to endeavor to use the libraries in a manner compatible to their intended design. Watch the walk through s that sound good, and copy what they do.

Mike Verta has a pretty good demonstrating his mod wheel technique, which is more aggressive than that used by many others, but you get the idea.