When you adjust the volume of a legato patch, for example, you're changing the volume of a single sample that was played at a specific dynamic. When you adjust the velocity while velocity X-fade is enabled, you're not only changing the volume, but you're also gradually fading between multiple patches, each played at a different, specific dynamic. This yields a tremendously realistic result since the timbre of the instrument now changes along with the volume, as it does with real instruments. When you go this route, you have absolute flexibility with regard to control over the length, dynamic, etc. of a given note, but there is the minor drawback that a small amount of realism is lost in the process of fading between samples. After all, the number of velocity layers at which a sample was actually recorded is a discrete number (usually 2, 3, or 4). Velocity X-fade can be particularly conspicuous when fading between dynamics with solo instruments, as you can sometimes hear two samples playing simultaneously.
The dynamic samples offer an alternative to the above approach. Each sample is a performance of a specific note with a specific duration (e.g., 2, 3, or 4 seconds) over which a crescendo or diminuendo (or combination thereof) occurs. There are usually a couple velocity layers for each sample, so one of the layers will be played for dynamics between 0 and 88, and the other will be played between 89 and 127, for example. Since each sample is an actual performance of a change in dynamics, these patches sound much more realistic than when static patches are varied in velocity. However, this realism comes at a price. Specifically, each instrument usually has less than a dozen or so different types of dynamic patches, so you have much less flexibility over the types of dynamic changes you can produce and the lengths of notes for which you can use them (though this problem can be overcome to a great extent with the time-stretch feature of VI Pro). Also, the samples naturally do not include the note transitions of the legato samples.
Here are a couple of examples of the differences between these approaches:
Dynamic patches vs. sustain w/ X-fade
The first pair of notes are Orchestral Violins, and the second pair of notes are from the Solo Viola. The first note of each pair is a dynamic crescendo-diminuendo patch, and the second note is an attempt to recreate the dynamic sound in a sustain patch using velocity X-fade and expression curves. While the Orchestral Violin samples sound pretty similar to me, the viola example is more difficult to reproduce with CC curves (notwithstanding the audible bow change, which I actually think sounds fine). Other people will be able to do a much better job of this than I, but it should give you an idea of how the two approaches differ.