I went back 3 years and found the average standard deviation of all players that played at least 60 games in each of the past 3 seasons. The average standard deviation for A1s was 0.16, and for A2s was 0.13. Not a huge difference, but it suggests that guys who put up big A1 totals are more likely to decrease their assist totals than those with high A2 totals. Counter intuitive, but it's what the numbers show. Even looking at the 25 best assist men over that 3 year span, the difference in standard deviation is even more pronounced (0.23 to 0.16).
Guys like Daniel Sedin, David Krejci and Alex Ovechkin have huge variations from season to season. Daniel and Ovi in particular have had a huge decline in primary assists while their secondary assists have not seen the same plunge.
This is of course highly dependent on line mates, so a guy that goes from playing with superstars to scrubs, or vice versa, will not follow the patterns of the average player. I would imagine that A1s correlate to talent and drop with age, while A2s correlate to the players you're surrounded by and are less affected by age.