Omron counter weird-ity

This problem came up about a year ago, and no-one at Omron support could figure it out, so I thought I'd throw it to the wolves.
No urgency in reply, of course, just looking for thoughts...

yesterday I had a oddity - Omron CPM1A register went funny
This is a machine where an AC drive moves a piece of sheet metal along a conveyor under three different cutters - spaced out linearly. The conveyor movement is read by an encoder.

I had a couple rungs give me an hours headache:

LD 1401 @MOV DM50 DM51 ;load a preset number eg 5 into a countdown register
LD 20000 AND 1005 @SUB DM51 #1 DM51 ;countdown when in AUTO and cutter fires
...
(do something to request cutter) ...SET 20005 ;latch cutter request
...
(cutter routine)
LD 20005 DIFU 20105 ;one shot for cutter request
LD 20105 SET 20205 ;latch it in
LD 20205 TIM5 #3
LD 20205 AND NOT TIM5 OUT 1005 ;fire cutter for 1/2 sec
LD TIM5 RSET 20205 RSET 20005 ;done with cutter

worked beautifully, as it should.

Then I got the brilliant idea that I should add a timer before the cutter fires, to give the conveyor time to stop:

the code sample becomes:
LD 20005 TIM13 #5 ;1/2 sec delay for stopping
LD TIM13 DIFU 20105 ;one shot for cutter request
LD 20105 SET 20205 ;latch it in
LD 20205 TIM5 #3
LD 20205 AND NOT TIM5 OUT 1005 ;fire cutter for 1/2 sec
LD TIM5 RSET 20205 RSET 20005 ;done with cutter

this identical segment was added to 2 different punches, worked perfectly, as it should.

BUT, adding this timer to this particular segment (I removed it and problem would go away) caused the DM51 countdown to decrement by 2 instead of 1 !

I though I must be getting a double action from the cutter or something, so I added this:

LD 1401 @MOV DM50 DM51 ;load a preset number eg 5 into a countdown register

LD 20000 AND 1005 @SUB DM51 #1 DM51 ;countdown when in AUTO and cutter fires
CNT5 DM50
LD 1401 (to the counter reset input)
SO..., these 2 counters have the same input condition, right?

Now with the timer 13 still active, DM51 would dec by 2 and CNT5 would dec by 1!!

I solved it by just scrapping the SUB and doing a MOV CNT5 DM51, instead, but why do you reckon this happened???

Key-word: 20005 settings ?

but the SET/RSET are above and below the COUNT rung, so the bit should be solid when the count is scanned...

Counting conditions for:
LD 20000 AND 1005 @SUB DM51 #1 DM51
works in every scan when true, but for: CNT5 DM50
it is dynamic (front edge of true) for Counter CNT5 property.
Could it be like that ?

Is this working ?
LD TIM5 OUT 'Memory'

LD 1401 @MOV DM50 DM51 

CLC ?

LD 20000 AND 1005 @SUB DM51 #1 DM51 ;...

(do something to request cutter) ...SET 20005 ;latch cutter request

...

the code sample becomes:

LD 20005 TIM13 #5 ;1/2 sec delay for stopping

LD TIM13 DIFU 20105 ;one shot for cutter request

LD 20105 SET 20205 ;latch it in

LD 20205 TIM5 #3

LD 20205 AND NOT 'Memory' OUT 1005 ;fire cutter for 1/2 sec

LD 'Memory' RSET 20205 RSET 20005 ;done with cutter