I corrected a massive error in my first answer... there are 6 or 7? total massages /second @ 4800 IIRC and the function for printing for both the Adafruit '7735 controller and most if not all of the UTFT library is blocking thus one of the issues. If the NMEA 0183 serial data is decoded and and the print function then displays the data either an interrupt is required or a simple solution as I wrote in my first reply it is necessary to print the whole screen in setup() and then ONLY print the data (Most Important) returned in the loop() that has CHANGED. while the GPS sentences are short... they still call the isr and will make printing stall (and everything else) stop until the isr completes, in this case again IIRC ... that's about 500 mS.. the easy answer for all of this is a state machine IMO. The short answer for all of this is a is to make everything work around 1 second as an interrupt and keep the functions that are required from a flag in the ISR as short as possible. Since the TimeGPS++ library isn't dependent on the 'age' of the GPS sentence as a whole it will anywhere it can hear.. GPS data as indicated by valid checksums. It Will return accurate time. AS I said before my Skylabs SKM53 can be instructed to listen to the most useful for time/static location data which are GGA and RMC. I bring it up again to point out that it should be looked for on the Mfr's datasheet for the device you own as it will free up a large chunk of time.
I made these measurements with a Saleae logic analyzer last year and I'm working from memory... IHTH.
Doc