You have not explicitly specified the number of permutations (which are used to calculate the p-value). The default value of the permutations is 0. It is not possible to calculate a p-value by permutations when no permutations are used. Hence I believe that the authors of the package deliberately used such a number which is clearly not a probability, to show that clearly.

Cibersort code

cibersort <- function(sig_matrix, mixture_file, perm = 0, QN = TRUE){
  ..
  P <- perm #number of permutations
  ..
  pval <- 9999
  ..
  if (P > 0) {
      pval <- 1 - (which.min(abs(nulldist - mix_r)) / length(nulldist))
  }
  ..
}

So, you can see from the code that if perm==0 (default), the pval is left with the value of 9999