Sorry, could you more specific about what this "sentence mining" study inside of Pleco would look like and how it differs from what we offer now? Are you basically looking for something with the sentence on the front of the card - without hiding (but maybe with boldfacing?) the target word - and then the word and its definition revealed on the back of the card? This is possible in the current beta with a ton of hacking, it'll be possible in the next update with a relatively easy to configure custom test type (now that we've made those easier to configure), and we've been thinking about making it a built-in thing you can select by default since it seems to be an increasingly popular study method.
As far as more elaborate exports: to be honest the whole export system is going to exist in a kind of in-between state for a little while because it's eventually going to be connected to the sync service which is our next project after getting 4.0 out, so I'd be reluctant to add much to that until whatever later 4.x update adds sync support, but something we can think about eventually. It would not be too hard to write something to reconstruct embedded sentences from a JSON export, though (but I'd wait on that for now because the JSON format is changing a bit even between now and 4.0 final).