I don’t expect predictive text algorithms to use context, there’s not enough room in the typical mobile phone for that, but they could use counts, and if so, what corpus did they get their numbers from?
German T9 doesn’t have Spargel in its lexicon. In May, why would you send a text message about anything else?