I was curious to check out Gasteiger charges on the oxygens of methyl acetate, while trying to see whether partial charges may be a decent indicator of hydrogen bond acceptor strenght of oxygens. Using the charge plugin via msketch, "chg on explicit H" being checked on, while "take resonant structures" and "use mikrospecies" were checked off. That gave me a proud -0.21 for the charge of the ester O (the carbonyl cousin being at -0.26). So far, so good. Marvin is 5.3.8, freshly updated.
Next, I wanted to compare this with the charge of the O of the Phe-O-Me ether. Being a lazy creature (as you well know), I just sketched the second species in the same sketcher, nearby the already present methy acetate. Surprise - as a consequence, the ESTHER O of the former came back with a partial charge of only -0.15!! And... it's coherent: I delete the PheOMe, I come back to -0.21. However, the =O of the esther does NOT move (it's 0.26 throughout the game). Neither does the PheOMe oxygen seem to care whether there's an esther (or anything else) sketched in the window - it's -0.17.
Now, this is funny: I agree that pKa calculations may run into trouble if several molecules are in the sketcher window (albeit... you use topological distances only - aren't these "infinte" between atoms of disjoined graphs?). Gasteiger, however, only runs through bonds - so, where does this bizzare (trizarre?) effect come from?