Re: ...web based system took three to five times the work and time.....
I can believe that. Real world example:
I was working for this company in the mid 1980's using an application written in a 4GL called PRO-IV. (info on PRO-IV can be found here: http://en.wikipedia.org/wiki/PROIV and here: http://www.proiv.com/ )
If you check out the first link, under "History", you will note a comment about MDIS, it was a port (for MSDOS) they had created that is one of the two platforms involved in this example.
The *nix variant was running on a Convergent Technologies S-80 (20 MHz 68020 CPU) supporting 16 terminal users. Data entry operations on this system "flew". Dumping over 140,000 inventory items into a report (for export to a spreadsheet for external manipulation) was quick. Trying the same task under MSDOS on a PC with a 66MHz CPU took forever. And we could not figure out why. The data and the application were on the same machine with only ONE user!
Finally, we made a call, and got the answer - the MSDOS port was going up and down the TCP/IP stack for disk i/o, even when it was on the same machine. No wonder performance sucked. So, I would not doubt that some of your performance issues are network related, in the sense that the workstation must pass the request to the server, wait for the server to chomp on the request, and spit out the result.
The *nix system's terminal screen's fixed text was drawn only once, on entry to the specific data entry screen; if you made multiple record entries, only the relevant data was moved down the wire. With the MSDOS system, the screen was redrawn each time it was changed, contributing to the sluggishness of the app. Needless to say, the shitty performance of a PC with this software convinced the owner to stick with the *nix box for quite a long time.
This points out one of the downsides of a interactive "web app" - everything (i.e. `the page`) has to be sent to the client each time something gets changed.