why umap showing bbknn perfectively integrated but tsne

Hi @all,
Thanks to develop the great tools,
I encounter a pecular problem on bbknn integrated data.
i follow the workflow code to run the data integrated,the code showing below,

 sc.pp.highly_variable_genes(adata)

adata = adata[:, adata.var['highly_variable']]
sc.tl.pca(adata, svd_solver='arpack')
sc.tl.tsne(adata)
///data integrated
sc.external.pp.bbknn(adata, batch_key='orig.ident')

sc.tl.umap(adata)
adata
sc.pl.umap(adata, color=['orig.ident'])
showing the well integrated, picture below,

But,when i want to see the tsne picture, the batch from different sample showing up on the tsne but umap like above picture
,i runing ,
sc.tl.tsne(adata)
sc.pl.tsne(adata, color=['orig.ident'])
the picture show below, indicating that the integrated can not be worked on tsne.

So, why this tsne showing significantly different with the object just running over the integrated process.
any advice would be appreciated
Best,
hanhuihong

Start by reviewing the workflow entry points named in the report: sc.tl.tsne, sc.external.pp.bbknn, sc.tl.umap, and the corresponding plotting calls. Check the documentation and implementation for how each method obtains its input representation; done means establishing whether the reported difference is expected or identifying a reproducible defect with a focused example.