Sliding past is usually more accurate.

The difference is, as you approach from the side, you have a solid egde that disrupts the inductive field more rapidly at close range, than if you approach end on where you have a solid lump that slowly disrupts the inductive field as it approaches from a distance.

However, if you want accuracy, I personally use optical slot sensors, as they're far less affected by temperature, and have a much smaller switching point, however they need to be reasonably well protected from contamination.
To give some theoretical figures, I just checked, and the last slot sensors I used are officially repeatable to 0.03mm over thier entire operating temperature, whereas having just checked the inductive sensor datasheet for the ones on my current lathe, and they list a +/-10% tolerance over their full operating temperature range, with switching distances of 8mm+/-10%, which gives a much wider switching point of 1.6mm +/-10%!
Off course, in practice repeatability is usually very good.

My current lathe uses inductive sensors, as they only provide a rough position which is then refined using the servo encoder index pulses to give highly repeatable homing (0.6 microns if my memory and calculations are correct!)

If you want to waste 15minutes, check out thise thread - Accuracy of Homing Devices? Pro's and Con's of Different Sensors
It has a couple links to other tests, and towards the end I helped Dickeybird setup accurate homing using a microswitch combined with a slotted disc and slot sensor.