Random effects perfectly correlated? (nlme package in R)

Figure I'm trying to simulate data and then fit a non-linear mixed effects model to it. But in most of the cases my random effects are perfectly correlated, and I don't understand why.And I think this is the reason for the frequently occuring errors (singularity of matrices or the algorithm doesn't converge...)

The figure shows the pairs plot for the random effects. The output of the model is given by

    [Nonlinear mixed-effects model fit by maximum likelihood
  Model: conc ~ SSmicmen(time, Vm, K) 
  Data: groupedData(conc ~ time | subject, data = sample2) 
  Log-likelihood: -101.3446
  Fixed: Vm + K ~ 1 
      Vm        K 
4.253410 8.732609 

Random effects:
 Formula: list(Vm ~ 1, K ~ 1)
 Level: subject
 Structure: General positive-definite, Log-Cholesky parametrization
         StdDev    Corr
Vm       1.1218405 Vm  
K        4.1308831 1   
Residual 0.9296998     

Number of Observations: 72
Number of Groups: 12][1] 

I'm new to NLMEM but this is not a good fit, if there is such a big correlation...? I would not want to reduce the random effects. I tried using different covariance matrices for generating the random effects, diagonal, non-diagonal... always the same problem. Can someone give me some explanations? Perhaps the sample size is not big enough? Thanks!