it's a good idea from Ivica to distinguish the different tasks on that topic discussing:
1. the implementation of users distribution based on Manhattan grid
2. the (available or to implement in the future in SEAMCAT) propagation model options for indoor-to-outdoor links, indoor-to-indoor links (within the same building) and indoor-to-indoor links (between two different buildings).
I also agree with Jose on the fact that points 1 & 2 can be addressed in parallel as they are independant.
Concerning the point 1., I see one challenging step consisting in trying to determine "easily" if possible (in the sense it does not require a lot of time to compute this information when running SEAMCAT) the intersection area between the Manhattan Grid indoor/outdoor area and the coverage/simulation area of a given device (Tx/Rx of the link) to generate the locations of the device based on the percentage of indoor/outdoor users selected by the SEAMCAT user in the Scenario.
The reason why I think it's not an obvious task is that the distribution of locations users should respect the selected indoor/outdoor users apportionment in any given scenario (if this apportionment is possible, e.g. if you select 100% outdoor/0% indoor and the device is located within a small area included in a building, it won't be possible unless you're considering user in the balcony !) while accounting the Manhattan grid topology.
The Manhattan Grid, as a 2D matrix, can be seen as a (mathematical) matrix of tiles with a given status: indoor (if the tile hosts a building) or outdoor (likely a street, not an open area which would be different for a Madrid grid). Now, one idea could be to determine the area (let's call A) related to the device on the Grid, based on the selected option, e.g. a radius between (Tx,Rx) or a set of parameters, e.g. a distance & an azimuth angle from another device. Therefore, each tile of A could belong to one of the two sub-areas of A: Aindoor & Aoutdoor. The location of the users can be generated , e.g. p_indoor percentage within Aindoor & p_outdoor percentage Aoutdoor. The tough part of this process seems to be the generic determination of A area.
If somebody has a different view (e.g. another solution easier to implement) or does not understand my explanations, don't hesitate to let me know.
Concerning point 2, I have a (Matlab) implementation of a part of the Recommendation ITU-R P.1238-8 available (attached to this message) that was done by one of my students some years ago. This needs to be polished, checked and improved but this could be a working baseline file for those who think this propagation model can be useful as a new PMP for SEAMCAT (as an alternative to one option of the Hata Model).