There are lots of good uses for incremental programming, and you should keep an open mind about these things.
In the olden days, all motions were incremental because the primitive NC "tape" machines would just read an incremental XYZ command from the paper tape, load the number in a register, and count up or down the number of pulses to make the move (probably with pulse motors). When microprocessors and mini-computers were introduced in what we now call "CNC" machines, the controls could keep track of the absolute position of each axis with no problem. Absolute or incremental moves could be made at any time.
I've seen incremental motions used in all kinds of situations, but here's one that few people know about: It involves using G91 (incremental) combined with the old method of uing non-decimal formatted numbers.
Suppose you're cutting a 3D mold with a ballnose endmill, and you're making a zillion little tiny moves in XYZ. The size of the program becomes a problem because these programs need to be fed to the CNC in "drip-feed" DNC mode. If you're running a DNC link at 9600 baud (960 characters/second), which program do you think will run faster:
This one (absolute) ?
X1.2345 Y2.3456 Z3.4567
X1.2347 Y2.3459 Z3.4571
X1.2350 Y2.3462 Z3.4577
X1.2355 Y2.3469 Z3.4579
.. or this one (incremental) ?
A lot smaller, isn't it?