No, the lengths are always the absolute length (they do not substract the fixed or hard-coded part).
This would make more sense:
... (in general!) the main problem with such masks is the fixed part of the mask, which could in theory affect the performance a lot (hashcat changes the left part most frequently in the kernel and if it is fixed, it can't change much... very easily expressed, simplified).
You could work around speed problems by testing --slow-candidates and/or pipes and or -a 6 (but again, this is more important for fast hash types, like MD4/MD5, for all only-pure - without optimized kernels - hash types it's not that significant.... but again, speed is king... therefore you just test and take the fastest method for cracking)
This would make more sense:
Code:
hashcat64.exe -a 3 -m 13400 --increment --increment-min=17 -1 ?l?u?d Keepass2.hash Sdkdjso38sLfj9ar?1?1?1?1?1?1?1?1?1... (in general!) the main problem with such masks is the fixed part of the mask, which could in theory affect the performance a lot (hashcat changes the left part most frequently in the kernel and if it is fixed, it can't change much... very easily expressed, simplified).
You could work around speed problems by testing --slow-candidates and/or pipes and or -a 6 (but again, this is more important for fast hash types, like MD4/MD5, for all only-pure - without optimized kernels - hash types it's not that significant.... but again, speed is king... therefore you just test and take the fastest method for cracking)