Elon previously touted end-to-end as being relatively simple because the tanker after attaching could impart a velocity change such that the liquids could flow without a pump.
A few plausible pieces of the pie:
1. Its really acceleration that's doing the lion's share of moving the fluids in the no-pump scenario, not velocity. So for the entire duration of the load transfer the vehicles would need to ~constantly be firing thrusters, otherwise its going to be a pretty slow and ineffective transfer.
2. The orientation of the vehicles is largely irrelevant for the no-pump concept--the vehicles can simply rotate to an orientation that aligns their velocity/acceleration vector opposite the desired direction of fluid flow.
3. There's probably some upside to having valves higher up on the vehicle (as opposed to just using the fill/drains on the aft ends). That's one plausible explanation for them being represented as joined "leeward to leeward". (You don't want to put
any hole let alone one for propellants on the re-entry side...)
4. There's some pressurization in the tanks (to mitigate cavitation and pick-up concerns); an active pump can better overcome unfavorable pressures.
5. Differences in molecular characteristics between the fuel and ox (specific gravities, etc.) make a no-pump configuration kind of complicated to get them both right.
6. Sort of a mash up of a few of above, there's an entropy aspect--with no forcing function (pump, vehicle acceleration), generally the propellant loads want to find a happy medium between the combined tanks. That's of course favorable when the target tanks are empty, but when the target tanks really starts to fill up, the liquids will naturally want to go back to the now-less-full tanker (and will make it increasingly harder to keep filling the target tanks).
7. Mostly 'cause some old school concepts are still cool, a paired vehicle set pitched down such that the engines are pointing anti-nadir is the most favorable gravity gradient orientation.