The reason, that there are so big differences between 1232 counter-parts, is that ciruit is much more complex, than people think. If someone had already tried to build such a circuit by discretes, he will believe me. There are so many hidden parameters, which you will never find in datasheet!
How complex internal circuitry is, you can imagine, when you have a look at datasheet of other supervisory chips. Sometimes they tell much more ybout internal circuitry...

One parameter, which is extremely important is delay of triggering after brown-out situation has been recognized. This is extremely important, because it determines, whether simple noise and glitches on Vcc is triggering 1232, or only a real brown-out condition does. If you have a look at datasheet of MAX690, e.g. you will find specification about 'maximum transient duration versus reset comparator overdrive'. And you will see, that MAX690 for 15mV overdrive of reset comparatar will need a maximum transient duration of 100µsec. For 30mV it's 50µsec and for 200mV it's 10µsec. This specification can mean a big advantage, because now noise and fast glitches on Vcc are no longer able to trigger brown-out detetcor. MAXIM has done this improvement, BECAUSE older and other supervisory chips showed trouble like that one Erik experienced. I don't think, that a DS1232 e.g. shows such a performance, because it is much older!
If a competitor like microchip is now manufacturing a pendant to DS1232, they have the choose: Being fully compatible to old industry standard DS1232 or to fabricate something different, something better perhaps. Both ways can be advantageous.

If you have a look with a high speed oscilloscope at a digital circuit consisting of some 74HCMOS chips you will notice noise and glitches on Vcc going well below threshold level, all the time! So, this delay of triggering is not only an important parameter, it is THE KEY PARAMETER. For me it's absolutely clear, that whenever I use such supervisory chip I do some low pass filtering on Vcc terminal of this chip. It's very easy to improve performance by just inserting a small resistor in Vcc line. 2R2 in combination with 100nF gives time constant of 220nsec, which is already good enough for most cases: Fast glitches are no longer able to make trouble. Of course, filtering can be further increased. Try 2R2 and 10µF tantal. This will give you a time constant in the range of 22µsec.

Don't make the mistake to think about a supervisory chip as being a digital component. Comparator stuff should be treated like analog circuitry, which allways needs some filtering.

Bye,
Kai