My understanding is that the Alfa system (on the Phase 2 at least) has only one point of measurement - a combined hot film flow sensor, and a temperature sensor mounted in the MAF sensor body. The hot film flow sensor is heated to a pre-determined temp, and subsequently cooled by the flow of inlet air. The amount the element is cooled is compared to the reference measurement provided by the temp sensor, and the mass of incoming air worked out by the ECU. Using this data along with throttle position, engine speed etc., the correct amount of fuel can be injected. Post combustion, the lambda probe measures the amount of oxygen remaining in the exhaust gasses, (an indication of the correct fuel/air ratio) and this data is added to the incoming air mass data to try and ensure the correct stoichiometric ratio, an important issue on cars with catalytic converters where too much unburnt fuel can damage the converter.
Its a great system in theory, if a little crude by today's standards, and fairly simple...which is why I am puzzled by the lean running I am experiencing!
Broadly speaking, the principles are the same. In the system you describe however, the degree of cooling is referenced against an embedded value of the element, whereas the one I described, the two temperature dependent components are selected (matched) and accuracy is thus much greater. How well the Alfa one works is very much dependent upon manufacture tolerances. There may even be a "Cold Junction" element within the E.C.U. to take into account the ambient temperature of the air. I.E., is the car being used in the Arctic or the Sahara. With the system I described, it doesn't matter. With the one you describe, it does.
However, what is crucial to the function of both methods is the linearity of the air flow over the sensor. The "Tubular Grid" I described is probably not an accurate enough description. It is a "Tubular Capillary Grid" about 150 mm long, which divides the air flow up such that any turbulence is greatly reduced to no more than the maximum diameter of any one capillary. This produces a uniform pressure front across the width of the orifice. The pressure rate changes due to mass air flow changes but it is a uniform rate of change, across the whole of the orifice.
All the MAF's I have seen, have a plastic grid across the face and it is largely assumed this is to protect the engine from debris. But the other function is to try to eliminate turbulence, but only when used in conjunction with the long ductwork. The Alfa manifold, along with every other manufacturers manifolds ( to my knowledge) is long as this greatly reduces turbulence and an element of "Ram Air" further improves the "Smoothing" effect. Modulation of the throttle also produces turbulence so this is also reduced by the long duct work and the element of Ram Air. Air flow systems and venturi are complex design areas. If the air flow pressure is greater at the edges of the MAF body than it is at the centre, where the flow is sampled, then the ECU will choose a fuel - air ratio (along with the other parameters) lower than the optimum, and you will get a lean burn mixture. Conversely, if the pressure is lower at the edges of the MAF than the center, where the air flow is sampled, you will get a rich mixture, resulting in sooting/caking of the Lambda Probes and as you rightly say, damage to the CAT.
It is surprising how air flow pressure can vary greatly across small areas such as throttle bodies/inlet manifolds and the mixture ratios we are considering are also very small. but the difference it can make to the overall performance of an engine is huge. Air flow inertia, just as oil flow inertia is an important factor so a lot goes into the design of such systems.
I cannot stress enough the need to go back to basics. Alfa is not my flavour of the month right now. But with regard to the original inlet manifold, I don't think they got it wrong. Probably because they didn't design it.