Why can't cuffdiff identify splicing difference

[Jolin]

I made simulation RNA-Seq data of one gene by FLUX. This gene has two isoforms. The ratio of the isoforms in the first sample is 0:100, while in the second sample 100:0. I ran the cufflinks, cuffcompare and cuffdiff, but the "splicing.diff" produced by cuffdiff showed that no significant splicing were found. But according to the "isoform_exp.diff", the FPKM of these two isoforms vary a lot in two samples. Why is the "test_stat" so small and why can't cuffdiff identify the splicing difference of this gene?

The content of splicing.diff is as follows:
test_id gene_id gene locus sample_1 sample_2 status value_1 value_2 sqrt(JS) test_stat p_value q_value significant
TSS1 XLOC_000001 A1_2_6 chr:4900-11900 test1 test2 OK 0 0 0.746756 2.07E-73 1 1 no

The content of isoform_exp.diff is as follows:
test_id gene_id gene locus sample_1 sample_2 status value_1 value_2 ln(fold_change) test_stat p_value q_value significant
TCONS_00000001 XLOC_000001 A1_2_6 chr:4900-11900 test1 test2 OK 261123 633.471 -6.02153 5.62055 1.90E-08 3.81E-08 yes
TCONS_00000002 XLOC_000001 A1_2_6 chr:4900-11900 test1 test2 OK 19398.3 300439 2.74006 -4.17333 3.00E-05 3.00E-05 yes