I think it is well worth the effort to convert the original sensor to read in PAR, and it is about as easy as it could be to do it. Once we know what the optimum set of filter squares is, to use the minimum number and have the best response curve, it is about a one hour job, at most. (I may make a few more calculations to try to find the best set of filters.)
Making a waterproof sensor is probably not worth the effort, and, while the last design I used was fairly easy to assemble, any mistake is very hard to correct, since it is all glued together. But, it does work. If you decide to do a waterproof sensor you almost have to use a disconnect fitting so you can use the original sensor to calibrate the new sensor. So, I think that is worth the effort too.