If we defer processing the touch events, pass the

original view coordinate with TouchEventData and use
them when we reprocess the events. We can't depend
on contentToView() to convert the doc point to the
view point. When we originally convert the view to
doc, the scroll position in UI doesn't change. When
we convert the doc back to view, we are chasing a
moving target. That is why we saw bigger delta.

Fix  http://b/issue?id=2577274