What are recommendations / best practices for using Expedition to generate routing in "complicated" environments and on short time scales?  By complicated I mean inshore, such as San Francisco Bay or perhaps Sydney Harbor where there are a lot of shoreline features, islands, etc.  Short time scales meaning a few hours.

I have been experimenting with creating marks and courses for buoy racing in San Francisco Bay and running routing solutions as a way to become proficient with Expedition and also to get insight into the strategy implications of current, wind direction, etc. From years of racing in the area I have a pretty good understanding of the currents and where to expect the shifts so often the results make sense but sometimes I have to scratch my head and think about the solution that the routing algorithm comes up with.  I'm not expecting a magic solution as it all depends on the accuracy of the wind data and the current model but it is interesting to look at and understand different solutions to hopefully gain some strategic insight about the course. 

My observations so far:
1. Isochronal routing doesn't work in these small scale (both distance and time) cases.  Perhaps if there's a way to change the time interval to on the order of 1 or 2 minutes, but I haven't found that yet.  Grid routing seems to work well.

2. Race Notes are important to keep the routing solution out of specific areas.  The Avoid Land feature (or high resolution option) doesn't always pick up the small features, my guess is this is due to the chart data.  Easy to deal with though.

3. Setting a mark in a course as "Leave to Port" or "Leave to Starboard" seems to pull the routing close to the mark, even in cases where it doesn't matter if the routing passes the mark within a few meters or 1000 meters, so long as it is left to the proper side. 

Probably these are newbie mistakes and there's something I'm missing. So far it seems to me that there's the offshore long distance racing use case where routing based on weather is important, and the inshore short distance/time use case where it's all about performance and tactical data. What I'm doing may not be a realistic use case. Any thoughts or recommendations?