I don't thing so, the problem must lie in your sample.
e.g., if your sample is just 2 values, there's a higher probability to your correlation be 1 or -1; although if your sample is composed by 4 variables, there will be a lesser probability to be 1 or -1... Think about it.
Also, you can easily perform some experiments using R. Create two variables V1 and V2, you can use rnorm(n) function to create a vector with n positions. After that, use cor(V1,V2) to get the correlation. Finally extract some samples and see what happens.