I have a theoretical question - what would happen in a UMAP embedding of 3 datasets with variable quality of the same sample type. Two of these datasets were generated using the 10x platform, and another using SmartSeq2. Below are the metrics of the datasets.

Dataset        batch           num cells    med. umi/cell    med. gene/cell
              10X REP1             4597        12411         3217
              10X REP2             5111         7121         1985
              SmartSeq REP1        1239       1550093       11305

It is clear that the SmartSeq sample has captured a lot more information compared to the two 10X replicates.

I can imagine that in a UMAP embedding of these datasets, the SmartSeq sample would form defined clusters, whereas the two 10x replicates (owing to a limitation in the data) may form false clusters that do not overlap with the Smart-Seq sample. Am I right in thinking this?