"Fake" and "real" are determined by lots of factors...
Say you're playing a flute instrument. The samples are of a real player who has a certain style of playing on his/her particular flute. The rate and amount of vibrato in the notes, the attack characteristics, etc. are all specific to that performer. How the producer of the sample recording session directed the player to perform also has a bearing on the vibe of the samples you're playing. So... is the vibe of that flute the right one for your track? Maybe, maybe not. It's a crapshoot because there is no such thing as "generic flute". There is [i]that[/i] flute that you're playing at the moment. So if your flute sounds fake, it could be because it's not the right flute "vibe" for the job. Might be time to find different flute samples.
Or... it could be because there weren't enough levels of dynamics recorded (p, mp, mf, f, ff) so you can't get enough levels of expression out of the instrument. Let's say the flute instrument has only one level of samples (no velocity switching or crossfading). When you play notes harder and softer you're simply not able to replicate the dynamics that you'd expect to hear a real flute player play with. Ergo, it might sound fake.
Now, there are lots of programming tricks that you can pull off to make even one-dynamic-level samples sound much more real and expressive, but I don't have time to get into that. That's a whole tutorial in an of itself. But save to say that if you play a note and expect the sample to do all the work, you're not going to get realism.
But even if the instrument does contain a lot of different dynamic levels and you wanted to make a note crescendo from soft to loud, the sound has to be programmed to let you smoothly crossfade from the soft samples through the loud samples, all while you're holding down a single key. In some instruments that's accomplished by holding down a note and moving a controller, like the modwheel. You absolutely cannot get that kind of smooth crescendo effect with velocity switching because it would mean repeating the note to increase the dynamics.
Two more things...
If you're not playing an instrument sound in a way that's idiomatic to the instrument, it's gonna sound fake. For example, if you try and play a fist-full of french horn notes, it's going to sound like cheese. Guaranteed.
Release time... if you don't tweak the release time of a sound to match the speed of your passages, it's gonna sound fake. Or maybe the release time is too short, so the ends of your notes sounds clipped. Adjustments to the release time are almost always needed to get the ends of notes not to sound like they're "swimming" or "keyboardy" or "choppy".
Finally, some sampled instruments just don't sound very good. Bad recordings, overprocessing, not enough samples taken (meaning that some pitches have to be stretched too far above/below their original pitch).
That's all I have time for, but if you want to know why some samples sound fake, there it is. It may not be the samples. It might be your approach to playing them. Or you could be using the wrong samples. Or you could be using the right samples and playing parts just perfectly but you need to "massage" your parts by adjusting release times, working to correct overlaps in the recorded MIDI notes and all kinds of other techniques that I don't have time to get into right now. Maybe later. But there are some of the reasons that samples can sound fake.
If you want help with a specific instrument sound, post back and we can discuss it further. And posting an MP3 or a link to a passage that doesn't sound real to you would help a ton, because talking about this in the abstract can only go so far.
Hope that gives you some perspective.