step down: high idle current, low efficency

[Max2018]

Thought I could re purpose existing Rohm BD9876AEFJ boards which are in a nutshell a copy of the schematic and layout from the datasheet.
datasheet: https://www.mouser.com/datasheet/2/348/rohm%20semiconductor_bd9876aefj-e-1201382.pdf
changes to the schematic
old config:
R1 = 10k + 100k wirewound poti (through hole)
R2 = 10k+ 10k
D1 = SD103A

new config:
R1 = 10k  + 100k wirewound poti (through hole)
R2 =3.92k + 0
D1 = SD103A

With the new config this IC consumes 90 mA @36 V  (3.24W!) without anything connected to it. With a fan connected as load it consumes 190mA (36 V) on the input with 170 mA output (24 V) which equals 60% efficiency ("just" 2,76W loss) while the datasheet suggest near 80% at this current.
As switching should have higher efficiency then a linear regulator I suspect somewhere an error. Could it be that the inductance in the feedback loop causes this issue?