ACS vs Census populations at the ZCTA level

Hi all

Just fyi, I downloaded
B01001, SEX BY AGE, Universe: Total population, 2008-2012 American Community Survey 5-Year Estimates
and compared it to
PCT3. SEX BY AGE. Universe: Total population. 2010 Census Summary File 2
both at the ZCTA level

Here is a comparison of the totals

ACS 2010 Census
State, Total 19,397,882 19,375,158
State, under 5 1,159,733 1,155,708

Pretty close you say? Well..... it is, at this level, but lets compare data from ZCTAs, looking at the category of under 5 (I use this age group a lot)
ZCTA .........ACS ..........2010 Census
10001........686.............624
10002........3271...........3620
10598........1260...........1481
11713........671.............702
11941........52...............98
12428........425............437
12429........0................11
12430........51..............67
13203........938...........1177
13208........2079.........1912
14456........952...........1042
14462.......18..............35
14464.......409............417

Well, in this non random sample, the numbers are generally in the same range at least. So if I'm just looking at population, at the ZCTA level, I'm not sure there is much advantage to using the ACS. Obviously, too, my sample isn't very random. At the state level, the ACS estimate is slightly larger than the 2010 Census estimate. But in most of the ZCTAs I chose (at least randomly to me), most of the 2010 Census estimates are larger than the ACS estimates.

I guess the question is, should I expect close correspondence between the ACS and Census 2010, and if not, why not.

Thanks

Gene
(sorry for the dots. I can't seem to get data tables lined up any other way)