Junction splicing

We’ve noticed many junctional splicing variants that scored very highly with AlphaGenome but somehow missed by SpliceAI. Of note, these are within the scanning window of SplicAI. Is this expected?