The reason for using one pole from each battery is to equalise the load seen by each battery. It's most important under high current discharge (like using a leccy winch) to equalise the discharge from the two batteries, and maximise available capacity. Under high current, the voltage drop of even a short cable (and often more importantly, the terminal joints on it) become easily measurable.
Now if you simply strap battery 2 to battery 1, and draw power from the terminals of battery 1, then the load seen by #2 is the load across #1 plus the resistance of two short cables and four extra terminations. Consequently, #2 will supply less than half the total current, and #1 a bit more than half.
If you use the +ve of #1 and the -ve of #2, then each battery sees the same load - #1 sees the load plus one extra cable (the -ve side one) and #2 sees the load plus one extra cable (the +ve one). Current delivered from each battery should be approximately equal.
The effect on total battery capacity delivered under high current is higher than might be expected because effective battery capacity is reduced under high load current. If this adds up to a significant 'real world' difference I don't know. You could calculate it fairly easily - google for 'Peukert constant' and grab a calculator. You need only an estimate for the resistance of a typical cable join.
Of course, the same is true (to a much smaller extent) when charging.
...TSD returns to lurking