Hi QAGB,
Thks yr comments. Taken on board. 
Have been attempting to slightly self-educate myself on this topic. I located the attached paper which is clearly not a recent publication (as are many other refs in this field) and may/may not relate to yr particular process style. You may be familiar with it already.
Liquid Sugar - counting methods.pdf 701.72KB
37 downloads
The product discussed is clearly of a much lower APC than you are producing so the “customized” analytical procedures are probably not relevant to yr situation however the comments regarding non-uniform distribution of microbial characteristics / sampling aspects could be of general interest. The significant question is whether the comments may relate to higher microbial levels as in yr product.
Just in passing, I also noted that some processors are routinely supplying product at very low APC levels such as <200cfu/g in this attachment –
Liquid_Sucrose.pdf 501.11KB
26 downloads
I noticed this textbook comment -
A substantial portion of the refinery's output is sold as liquid sugar. This can be in the form of (1) dissolved granulated sugar, (2) fine liquor or (3) intermediate liquors from the various crystallising stages.
I'm only speculating via the limits you previously gave but I guess 1 > 3 may represent a (substantially) increasing progression in APC value (and similarly in max. limit values). > 10:1 perhaps ? (not too sure what "fine" liquor actually is)
The product currently under discussion is perhaps from "3" ?
I don't know how you are determining / applying the implemented heat treatment but as I understand, yr main problem is not that you cannot get to the desired low APC level, it's more that you cannot get there consistently due to the large swings in results for the ONE sample unit submitted per batch.
Frankly, assuming you apply some kind of xD logic to meet the APC limit, I find it difficult to see how you can plan any heat treatment without some assumption regarding the max. starting APC level. But perhaps you've found a way, eg Big Bazooka (BB). I do wonder how you convince yourselves that the APC results for successive batches should be the same ?.
I can think of a variety of possible causes to explain the variations you are getting, eg -
(1) different batches have significantly different start APCs, causes similar variations in output results - ie may be genuine reported variations if lab is reliable
(2) diferrent batches have similar start APCs but the applied heat treatment varies - ie may be genuine reported variations if lab is reliable
(3) there is a significant variation in APC within any given lot - need data
(4) there is a significant variation in APC between different lots - need data
(5) the calculated/applied heat treatment is not consistently correlated to the starting APC so cannot achieve a similar result - need method
(6) the sampling procedure for submitted sample is not aseptic - need method.
(7) the lab handling/analytical procedure is not consistent/correct - need to test with random duplicates
The classic statistical method to solve such problems, ie identify the weak point, is via methods like ANOVA but they entail a lot of work and analysis. Probably easiest route is to examine the most accessible elements first.
Personally i would validate the lab first. If Ok then the sampling. if OK then the process. Difficult to offer much thouight on the process due limited knowledge (2-ways).
One popular, albeit crude, solution is a BB approach, eg target low enough so that the swings do not give results out of compliance. This is the approach i am forced to use for packed weight control of some products where major texture variations within/between lots simply rule out close control. Problem is that it may cause a financial hiccup. Maybe you are doing this now ?
I also noticed a couple of sites with micro. sampling plans for liquid sugar based on lot sizes although i daresay you are already well informed on this subject. Can post if you are interested (and i can re-find).