SymEigsSolver misses 1 eigen value

[linaelf]

I used SymEigsSover to calculate k smallest eigen values as below code.

Spectra::SymEigsSolver<SparseSymMatProd<double>> eigs(op, k, ncv);
eigs.init();
int nconv = eigs.compute(SortRule::SmallestAlge, 1000, 1e-10, SortRule::SmallestAlge);

Here's the input matrix data: Matrix(834x834, delimited with space)

Seems diffrent k produces different eigen values for matrix. One value -1 is missing when k=20.
k=20, ncv=50, eigen values are:
-1
-1
-0.991601
-0.972826
-0.964733
-0.957468
-0.953686
-0.951441
-0.94711
-0.931136
-0.929913
-0.926608
-0.914001
-0.904855
-0.900097
-0.882461
-0.875527
-0.859567
-0.812831
-0.811643

k=25, ncv=50, eigen values are:
-1
-1
-1
-0.991601
-0.972826
-0.964733
-0.957468
-0.953686
-0.951441
-0.94711
-0.931136
-0.929913
-0.926608
-0.914001
-0.904855
-0.900097
-0.882461
-0.875527
-0.859567
-0.812831
-0.811643
-0.628877
-0.404174
-0.397855
-0.390249

[yixuan]

Sorry for being late, but it seems the link to the matrix file is now invalid. Could you share the file again?