I suspect if you were to compare code that uses both the remainder and quotient of a division, you would find a similar trend: the x86 division instructions produce both, but ARM has only a quotient-generating division instruction, and you have to do another multiplication(!) and subtraction to get the remainder.

That, along with the fact that earlier ARMs didn't even have a divide instruction, is one of the reasons why I'm not at all fan of the "RISC approach" --- division may not have been easy to implement, but microcoding would've been at least as fast as doing the algorithm in software while taking up far less space, and with hardware advances that move operations from microcode into direct execution, existing software that already has those instructions would automatically become faster on newer hardware with no other action needed.

Division hardware almost always will generate the remainder along with the quotient, but ARM would either need to add another instruction (no benefit to existing software, only new software that knows how to use it) or attempt to detect patterns of simpler instructions that are doing the same thing in order to replace them with one "fused uop" (much more difficult to do.)