Update: the folks at NHLBI have been following up very thoroughly. The hard CHD function has indeed been published and can be found here (at the very bottom):
http://www.framinghamheartstudy.org/risk/hrdcoronary.html

It seems the equation they use on their online calculator is more sophisticated than the point-based rule. Unfortunately the equation has not been published, fortunately the point-based system shows these discrepancies only with extreme values and the group (low, med, high) will most of the time be accurate. Here is their full response:

-----
The risk calculations associated with the Adult Treatment Panel III (ATP III) guidelines are based on a new updated version of the previous Framingham equation; therefore it will give different results than previous versions. Furthermore, the electronic versions of the ATP III risk-calculation products (i.e., the online risk calculator, spreadsheet, and Palm OS tool) use a different risk determination methodology than do the pen and pencil method of assessing 10-year risk (i.e., the ATP III Guidelines At-A Glance Quick Desk Reference and the ATP III Executive Summary).

The electronic products calculate risk using a complex mathematical formula based on Framingham data. This formula, which contains multiple natural logarithm and exponent terms, is too complicated to be calculated manually in the clinical setting. However, the point-counting system used in the pen and pencil method provides a way to manually calculate 10-year risk. The point-counting system approximates the risk calculation equation, but cannot do so perfectly, particularly with input of extremely high or low data values.

The ATP III electronic products include the following statement when presenting the results of risk calculations: “The risk score shown was derived on the basis of an equation. Other National Cholesterol Education Program materials, such as ATP III print products, use a point-based system to calculate a risk score that approximates the equation-based one.”

Consider, for example three 10-year risk scores: 8, 12, and 16 percent. The 8-percent 10-year risk score should be eliminated, as this formula has been updated. We then are then left with the 16-percent 10-year risk score from the electronic risk calculators, and the 12-percent 10-year risk from the pen and pencil method. Although these numbers are different, both numbers fall into the 10–20 percent 10-year risk category and have the same LDL goal of less than 130 mg/dL.

The ATP III Guidelines are guides, and not hard and fast rules. Clinical judgment, in addition to the patient’s current and past medical history, are factors to consider when prescribing treatment.
----

Thanks for chiming in. I sent them an email asking whether the mentioned equation is published and whether the equation or the discrete model would be authoritative.

Agree. Don't know that the formula for hard cardiac events that uses continuous rather than discrete variables has been published.

There is a big issue with the online calculator -- it doesn't use the point system devised in the final publication (http://circ.ahajournals.org/content/106/25/3143.full.pdf). This is a problem, the differences can be substantial (8% point based vs 12% from the online calculator), and of course people will assume MedCalc is wrong. Has the continuous model been published?

(34 yo male, 220 total chol, 35 HDL, 165 systole on medication, smoker: 8% point based vs. 12% from online tool)

Also should link results to USPSTF aspirin profylaxis rec for men which are based on 10 year cardiovascular risk

Zur Dokumentation in Praxis, Absicherung gegenüber Kassen Verordnung CSE Hemmer verwendet
nicht Ideal, jedoch üblich

Thanks, people being confused is a good point and one we should address!

Link to the nhlbi site works and leads to the continuous model. There are also discrete model calculators for Framingham risk out there which give results close to the continuous model and should be easy to incorporate.
Important to note that these are formulas for hard cardiac events. You currently have a Framingham calculator for soft cardiovascular events. The hard event rate is what is used in deciding whether to start cholesterol lowering agents and would be helpful. Some users may be confused as to which formula you are currently using.