I would like to know if we can rated HRC#3 480V panelboards

Not entirely clear on your question. But if you use the table method in NFPA 70E and are within the parameters of the table, the highest HRC is 2. However, from doing many arc flash analyzes, I have seen 480V panelboards with incident energy greater than 25 cal/cm^2 so I myself do not trust the tables to provide proper PPE

Thing is that we have 480V panelboards rated HRC#3 (the method used were the tables) and here is an issue if this shall be hRC#2 instead of 3. I'm not sure if we need to move forward in this direction or if we can keep the rated of 3 as well.

It has never been a 3. The assumptions (and calculations) have not changed for several years. My guess is that one of the following happened in the past:
1. Someone that had little or no knowledge just simply did something knee jerk and figured more protection is better and set a standard of 3.
2. Someone did an actual calculation of some sort and found the worst rating was equivalent to a "3".

It has been proven based on case histories collected at Dupont that the tables will specify adequate arc flash PPE 50% of the time. This is better than nothing but lless than the 95% accuracy estimated with IEEE 1584. The major issue with the current tables are believed to be where tasks are "derated" such as H/RC 1 when the incident energy is 2. 50% is not really good odds but it's better than the 85%+ chance of an injury requiring hospitalization from the same study without using any particular arc flash analysis method. Hopefully with the upcoming changes in 2015 where the "derating" goes away, so will the injury statistics.

The major fallacy here though is the assumptions on the tables which are based on a maximum trip time and a maximum short circuit current rating (and implicitly, some equipment design details). This is enough information to perform an incident energy calculation. However:
1. It is very easy to calculate short circuit current rating of a transformer assuming that the upstream side is a stiff bus (infinite current available) and that the downstream wire has zero impedance. In this case the only impedance that matters is the transformer itself. This method is commonly used for sizing breakers, panelboards, MCC's, etc.
2. It is not uncommon to compare this to the maximum rating in the arc flash table. In itself this is pretty innocent.
3. The mistake though is then looking up this value on a time-current-curve and using that value for a trip time. First off the fault current is unlikely to get that high. The breaker if sized this way may NEVER trip. Second this very conservative estimate results in a very fast (often instantaneous) trip time. In the real world currents are often significantly lower than this. Thus the trip time turns out to be much longer.
4. The resulting energy very often exceeds the assumptions made in the tables. Thus turns out to be true a significant portion of the time (45%?).

For these reasons I strongly discourage use of the tables. Use them when they don't have another option, but put all the non-crisis work load into fixing it.

X2 on what Paul said. You need to know the actual fault current and associated protective trip device at EACH panel to use the tables. By the time you have spent the time and money to do that, it is a very minor effort to calculate the arc flash. All the modeling needed to figure out the fault current and trip time at EACH location is the information needed for arc flash hazard analysis. There is also a host of other benefits of doing an actual study.

Like Paul said, there never was a HRC 3 in NFPA Tables for 480V panelboards even though from years of doing studies I have seen everything from <1 cal/cm^2 to over 40 cal/cm^2 on 480V panelboards.